   <Dialog file="">
        <Body>
            <Topics/>
            <Turn nickname="(#user2#)" genid="8">
                <Utterance genid="9" ref="-1" time="08:57:32" date="11/12/2007" oldid="8" color="" topic="">hello</Utterance>
                <Utterance genid="10" ref="-1" time="08:58:15" date="11/12/2007" oldid="9" color="" topic="">are you ready, guys?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="11">
                <Utterance genid="12" ref="-1" time="08:59:37" date="11/12/2007" oldid="10" color="" topic="">yes, i am ready</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="13">
                <Utterance genid="14" ref="-1" time="08:59:47" date="11/12/2007" oldid="11" color="" topic="">yes, i'm too</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="15">
                <Utterance genid="16" ref="-1" time="08:59:53" date="11/12/2007" oldid="12" color="" topic="">great</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="17">
                <Utterance genid="18" ref="-1" time="08:59:57" date="11/12/2007" oldid="13" color="" topic="">:)</Utterance>
                <Utterance genid="19" ref="-1" time="09:00:04" date="11/12/2007" oldid="14" color="" topic="">lets start</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="20">
                <Utterance genid="21" ref="-1" time="09:00:24" date="11/12/2007" oldid="15" color="" topic="">i propose each one of us to make a presentation of his method first</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="22">
                <Utterance genid="23" ref="21" time="09:00:43" date="11/12/2007" oldid="16" color="" topic="">this is a very good idea</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="24">
                <Utterance genid="25" ref="-1" time="09:00:50" date="11/12/2007" oldid="17" color="" topic="">:)</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="26">
                <Utterance genid="27" ref="-1" time="09:01:00" date="11/12/2007" oldid="18" color="" topic="">who wants to start first?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="28">
                <Utterance genid="29" ref="-1" time="09:01:07" date="11/12/2007" oldid="19" color="" topic="">who is the first?</Utterance>
                <Utterance genid="30" ref="-1" time="09:01:11" date="11/12/2007" oldid="20" color="" topic="">:)</Utterance>
                <Utterance genid="31" ref="-1" time="09:01:12" date="11/12/2007" oldid="21" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="32">
                <Utterance genid="33" ref="-1" time="09:01:25" date="11/12/2007" oldid="22" color="" topic="">i'll go, if you don't mind</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="34">
                <Utterance genid="35" ref="-1" time="09:01:32" date="11/12/2007" oldid="23" color="" topic="">yes, sure</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="36">
                <Utterance genid="37" ref="-1" time="09:01:35" date="11/12/2007" oldid="24" color="" topic="">i'm agree</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="38">
                <Utterance genid="39" ref="-1" time="09:02:07" date="11/12/2007" oldid="25" color="" topic="">i'm here to support the Naive Bayes classifier</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="40">
                <Utterance genid="41" ref="39" time="09:02:33" date="11/12/2007" oldid="26" color="" topic="">please describe in short your classifier</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="42">
                <Utterance genid="43" ref="39" time="09:03:52" date="11/12/2007" oldid="27" color="" topic="">do it</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="44">
                <Utterance genid="45" ref="-1" time="09:07:37" date="11/12/2007" oldid="28" color="" topic="">ok, i think that (#user2#) got disconnected, or something...</Utterance>
                <Utterance genid="46" ref="-1" time="09:07:48" date="11/12/2007" oldid="29" color="" topic="">let's hope he will recover after this..</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="47">
                <Utterance genid="48" ref="-1" time="09:08:03" date="11/12/2007" oldid="30" color="" topic="">so, the NB classifier is a very popular algorithm due to its simplicity, computational efficiency and its surprisingly good performance for real-world probllems</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="49">
                <Utterance genid="50" ref="-1" time="09:08:40" date="11/12/2007" oldid="31" color="" topic="">can you be more specific when you say real-world problems?</Utterance>
                <Utterance genid="51" ref="-1" time="09:09:22" date="11/12/2007" oldid="32" color="" topic="">can you present for as some data sets or data distributions that fits best the NB classifier</Utterance>
                <Utterance genid="52" ref="-1" time="09:12:13" date="11/12/2007" oldid="33" color="" topic="">do you need some help?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="53">
                <Utterance genid="54" ref="50" time="09:12:43" date="11/12/2007" oldid="34" color="" topic="">it is used in different domains, the most interesting for us is text classification, so the input data is text</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="55">
                <Utterance genid="56" ref="-1" time="09:13:22" date="11/12/2007" oldid="35" color="" topic="">do you meen speech recognition?</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="57">
                <Utterance genid="58" ref="56" time="09:14:39" date="11/12/2007" oldid="36" color="" topic="">i meant news classification in categories, e-mail classification in spam and non-spam</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="59">
                <Utterance genid="60" ref="-1" time="09:15:16" date="11/12/2007" oldid="37" color="" topic="">yes...you say text...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="61">
                <Utterance genid="62" ref="-1" time="09:15:45" date="11/12/2007" oldid="38" color="" topic="">basically it's an algorithm and it can be applied wherever it is needed</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="63">
                <Utterance genid="64" ref="-1" time="09:16:40" date="11/12/2007" oldid="39" color="" topic="">ok...describe your model...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="65">
                <Utterance genid="66" ref="-1" time="09:16:43" date="11/12/2007" oldid="40" color="" topic="">ok, i think this is the main role for all the classifers</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="67">
                <Utterance genid="68" ref="-1" time="09:17:00" date="11/12/2007" oldid="41" color="" topic="">so i go on ...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="69">
                <Utterance genid="70" ref="68" time="09:17:23" date="11/12/2007" oldid="42" color="" topic="">yes, you should continue</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="71">
                <Utterance genid="72" ref="-1" time="09:17:26" date="11/12/2007" oldid="43" color="" topic="">The naive attribute comes from the fact that the model assumes that all features that are classified are fully independent</Utterance>
                <Utterance genid="73" ref="-1" time="09:18:22" date="11/12/2007" oldid="44" color="" topic="">Even though this classifier is also known as Idiots Bayes and in spite of the assumptions and the simplistic design of the classifier, it turns out to be very well suited for problems involving normal distributions, which are very common in real-world problems</Utterance>
                <Utterance genid="74" ref="-1" time="09:18:42" date="11/12/2007" oldid="45" color="" topic="">It also often works much better in many complex real-world situations than one might expect</Utterance>
                <Utterance genid="75" ref="-1" time="09:19:20" date="11/12/2007" oldid="46" color="" topic="">So, a naive Bayes classifier is a simple probabilistic classifier based on applying Bayes' theorem. Because of the assumptions, a more descriptive term for the underlying probability model would be "independent feature model"</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="76">
                <Utterance genid="77" ref="74" time="09:19:44" date="11/12/2007" oldid="47" color="" topic="">ok, i got the picture</Utterance>
                <Utterance genid="78" ref="-1" time="09:20:44" date="11/12/2007" oldid="48" color="" topic="">i think we need to keep in mind that we are company directors, and each of our company should pormote one classifier</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="79">
                <Utterance genid="80" ref="-1" time="09:21:15" date="11/12/2007" oldid="49" color="" topic="">yes we must be in touch with scope</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="81">
                <Utterance genid="82" ref="-1" time="09:21:35" date="11/12/2007" oldid="50" color="" topic="">I propose that after each one will present his classifier, to discuss the pros and cons about why one should use one classifier and not the other</Utterance>
                <Utterance genid="83" ref="-1" time="09:22:16" date="11/12/2007" oldid="51" color="" topic="">at the end we should focus to try to find a real word example/distribution of input data that should use all the classifiers...</Utterance>
                <Utterance genid="84" ref="-1" time="09:22:46" date="11/12/2007" oldid="52" color="" topic="">I will talk about SVM - support vector machines</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="85">
                <Utterance genid="86" ref="82" time="09:22:55" date="11/12/2007" oldid="53" color="" topic="">to discuss this i sugest to speek about avantages and desavantages...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="87">
                <Utterance genid="88" ref="84" time="09:23:08" date="11/12/2007" oldid="54" color="" topic="">please go ahead</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="89">
                <Utterance genid="90" ref="-1" time="09:23:19" date="11/12/2007" oldid="55" color="" topic="">do you have anything to say before i start? because i will prove that this classifier is the best:d</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="91">
                <Utterance genid="92" ref="-1" time="09:23:33" date="11/12/2007" oldid="56" color="" topic="">one moment</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="93">
                <Utterance genid="94" ref="-1" time="09:23:38" date="11/12/2007" oldid="57" color="" topic="">ok.. shoot</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="95">
                <Utterance genid="96" ref="-1" time="09:23:53" date="11/12/2007" oldid="58" color="" topic="">(#user2#) say something about avantages of your model...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="97">
                <Utterance genid="98" ref="96" time="09:24:20" date="11/12/2007" oldid="59" color="" topic="">ok then</Utterance>
                <Utterance genid="99" ref="-1" time="09:24:42" date="11/12/2007" oldid="60" color="" topic="">the biggest advantages are:</Utterance>
                <Utterance genid="100" ref="-1" time="09:25:16" date="11/12/2007" oldid="61" color="" topic="">it requires a small amount of training data to estimate the parameters necessary for classification</Utterance>
                <Utterance genid="101" ref="-1" time="09:25:30" date="11/12/2007" oldid="62" color="" topic="">simple algorithm</Utterance>
                <Utterance genid="102" ref="-1" time="09:25:40" date="11/12/2007" oldid="63" color="" topic="">fast to train, fast to evaluate</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="103">
                <Utterance genid="104" ref="-1" time="09:25:44" date="11/12/2007" oldid="64" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="105">
                <Utterance genid="106" ref="-1" time="09:25:47" date="11/12/2007" oldid="65" color="" topic="">surprisingly good for real-world problems</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="107">
                <Utterance genid="108" ref="-1" time="09:26:07" date="11/12/2007" oldid="66" color="" topic="">ok.. i think it is enough for now, i will start my presentation</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="109">
                <Utterance genid="110" ref="-1" time="09:26:13" date="11/12/2007" oldid="67" color="" topic="">yes</Utterance>
                <Utterance genid="111" ref="-1" time="09:26:17" date="11/12/2007" oldid="68" color="" topic="">do it (#user1#)</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="112">
                <Utterance genid="113" ref="-1" time="09:26:27" date="11/12/2007" oldid="69" color="" topic="">SVMs are a set of related supervised learning methods used for classification and regression.</Utterance>
                <Utterance genid="114" ref="-1" time="09:27:17" date="11/12/2007" oldid="70" color="" topic="">they are linear classifiers and are based on discriminative data</Utterance>
                <Utterance genid="115" ref="-1" time="09:27:56" date="11/12/2007" oldid="71" color="" topic="">the main difference compared to the model that (#user2#) presented ( NB), SVMs isn't a model based classifier</Utterance>
                <Utterance genid="116" ref="-1" time="09:28:30" date="11/12/2007" oldid="72" color="" topic="">As a classifier, SVM it is used mostly for pattern recognition</Utterance>
                <Utterance genid="117" ref="-1" time="09:29:09" date="11/12/2007" oldid="73" color="" topic="">You can think at patterns as voice recognition, structure recognition, ADN recognition and so on...</Utterance>
                <Utterance genid="118" ref="-1" time="09:29:39" date="11/12/2007" oldid="74" color="" topic="">SVMs were used with success to classifie some cancer types.</Utterance>
                <Utterance genid="119" ref="-1" time="09:30:04" date="11/12/2007" oldid="75" color="" topic="">are you there? because you don't ask me anything...</Utterance>
                <Utterance genid="120" ref="-1" time="09:30:35" date="11/12/2007" oldid="76" color="" topic="">i don't think i am very good at presentations so you will understand everything and have nothing to ask</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="121">
                <Utterance genid="122" ref="119" time="09:30:37" date="11/12/2007" oldid="77" color="" topic="">yep, just following your interesting presentation</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="123">
                <Utterance genid="124" ref="-1" time="09:31:57" date="11/12/2007" oldid="78" color="" topic="">Then, let's go a little furher and have a picture about the mathematical background used by SVM</Utterance>
                <Utterance genid="125" ref="-1" time="09:32:05" date="11/12/2007" oldid="79" color="" topic="">are you ok with this?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="126">
                <Utterance genid="127" ref="-1" time="09:32:10" date="11/12/2007" oldid="80" color="" topic="">yes</Utterance>
                <Utterance genid="128" ref="-1" time="09:32:16" date="11/12/2007" oldid="81" color="" topic="">present it</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="129">
                <Utterance genid="130" ref="125" time="09:32:53" date="11/12/2007" oldid="82" color="" topic="">yes, but not please just in general</Utterance>
                <Utterance genid="131" ref="130" time="09:33:15" date="11/12/2007" oldid="83" color="" topic="">without not</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="132">
                <Utterance genid="133" ref="-1" time="09:33:16" date="11/12/2007" oldid="84" color="" topic="">the name for Support vector machines comes from the fact that the mathematical background for SVM are ... as you might guess some multi-dimensional vectors</Utterance>
                <Utterance genid="134" ref="-1" time="09:33:59" date="11/12/2007" oldid="85" color="" topic="">To classifie data SVM must be trained as part of a machine-learning process.</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="135">
                <Utterance genid="136" ref="134" time="09:34:53" date="11/12/2007" oldid="86" color="" topic="">so it is a supervized learning method, isn't it?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="137">
                <Utterance genid="138" ref="-1" time="09:35:07" date="11/12/2007" oldid="87" color="" topic="">Based on some input data, the SVM will be trained to classifie new data. The only restriction is that the data that will be classified to have the same distribution as the data for which SVM was trained for</Utterance>
                <Utterance genid="139" ref="136" time="09:36:06" date="11/12/2007" oldid="88" color="" topic="">yes, of course</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="140">
                <Utterance genid="141" ref="-1" time="09:36:11" date="11/12/2007" oldid="89" color="" topic="">so...it not be used in case different distribution for the same problem?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="142">
                <Utterance genid="143" ref="-1" time="09:36:36" date="11/12/2007" oldid="90" color="" topic="">not, the error cannot be estimated in this case</Utterance>
                <Utterance genid="144" ref="-1" time="09:36:53" date="11/12/2007" oldid="91" color="" topic="">so, it is a good practice to train on the same type of data</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="145">
                <Utterance genid="146" ref="134" time="09:37:57" date="11/12/2007" oldid="92" color="" topic="">i believe all 3 methods are used for learning some given data first, and then as a classifier for test data</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="147">
                <Utterance genid="148" ref="-1" time="09:38:38" date="11/12/2007" oldid="93" color="" topic="">yeap, this is sure, because you cannot classifie new data until you don't have something already learned...</Utterance>
                <Utterance genid="149" ref="-1" time="09:39:00" date="11/12/2007" oldid="94" color="" topic="">*classify</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="150">
                <Utterance genid="151" ref="148" time="09:39:26" date="11/12/2007" oldid="95" color="" topic="">this is the difference between classification and clustering</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="152">
                <Utterance genid="153" ref="-1" time="09:39:33" date="11/12/2007" oldid="96" color="" topic="">sorry about my typing</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="154">
                <Utterance genid="155" ref="153" time="09:39:54" date="11/12/2007" oldid="97" color="" topic="">no problem, we're all tired :)</Utterance>
                <Utterance genid="156" ref="-1" time="09:40:15" date="11/12/2007" oldid="98" color="" topic="">so what are the main advantages of SVM's?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="157">
                <Utterance genid="158" ref="-1" time="09:40:39" date="11/12/2007" oldid="99" color="" topic="">so , the vectors have some components that place the problem into multispace dimmension</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="159">
                <Utterance genid="160" ref="146" time="09:41:24" date="11/12/2007" oldid="100" color="" topic="">the problem consist in the best method of classification...so we have some new dates....</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="161">
                <Utterance genid="162" ref="-1" time="09:41:25" date="11/12/2007" oldid="101" color="" topic="">hmm, i think i want to make a general presentation and you both know what i am trying to present ...</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="163">
                <Utterance genid="164" ref="-1" time="09:41:36" date="11/12/2007" oldid="102" color="" topic="">ok</Utterance>
                <Utterance genid="165" ref="-1" time="09:41:37" date="11/12/2007" oldid="103" color="" topic="">do it</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="166">
                <Utterance genid="167" ref="-1" time="09:42:54" date="11/12/2007" oldid="104" color="" topic="">the main advantage of the SVMs is the fact that classification is made based on some maths formula...</Utterance>
                <Utterance genid="168" ref="-1" time="09:43:18" date="11/12/2007" oldid="105" color="" topic="">and it is very easy to extend on problem with a new dimension</Utterance>
                <Utterance genid="169" ref="-1" time="09:43:49" date="11/12/2007" oldid="106" color="" topic="">the algorithm remains, you just need to add a new dimension to the problem</Utterance>
                <Utterance genid="170" ref="-1" time="09:44:13" date="11/12/2007" oldid="107" color="" topic="">let's look at the whiteboard</Utterance>
                <Utterance genid="171" ref="-1" time="09:45:35" date="11/12/2007" oldid="108" color="" topic="">we have 2 kind of data</Utterance>
                <Utterance genid="172" ref="-1" time="09:46:32" date="11/12/2007" oldid="109" color="" topic="">and when we received an input data that needs to be classified, we must determine to which class this new data belongs....</Utterance>
                <Utterance genid="173" ref="-1" time="09:47:26" date="11/12/2007" oldid="110" color="" topic="">for this, we determining the plane that separe the 2 data</Utterance>
                <Utterance genid="174" ref="-1" time="09:47:28" date="11/12/2007" oldid="111" color="" topic="">types</Utterance>
                <Utterance genid="175" ref="-1" time="09:48:02" date="11/12/2007" oldid="112" color="" topic="">There are many linear classifiers that might satisfy this property. However, we are additionally interested in finding out if we can achieve maximum separation (margin) between the two classes</Utterance>
                <Utterance genid="176" ref="-1" time="09:48:10" date="11/12/2007" oldid="113" color="" topic="">By this we mean that we pick the hyperplane so that the distance from the hyperplane to the nearest data point is maximized.</Utterance>
                <Utterance genid="177" ref="-1" time="09:48:27" date="11/12/2007" oldid="114" color="" topic="">in the presentation above, a hyperplane is a plane for multidimension data</Utterance>
                <Utterance genid="178" ref="-1" time="09:48:38" date="11/12/2007" oldid="115" color="" topic="">in my example the dimension of the data is 2</Utterance>
                <Utterance genid="179" ref="-1" time="09:49:01" date="11/12/2007" oldid="116" color="" topic="">so, we need that nearest distance between a point in one separated hyperplane and a point in the other separated hyperplane is maximized.</Utterance>
                <Utterance genid="180" ref="-1" time="09:49:23" date="11/12/2007" oldid="117" color="" topic="">This (hyper)plane known as the maximum-margin hyperplane and such a linear classifier is known as a maximum margin classifier</Utterance>
                <Utterance genid="181" ref="-1" time="09:49:36" date="11/12/2007" oldid="118" color="" topic="">i will try to be short</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="182">
                <Utterance genid="183" ref="-1" time="09:49:39" date="11/12/2007" oldid="119" color="" topic="">ok...if we have some data 1 in data2...</Utterance>
                <Utterance genid="184" ref="-1" time="09:49:53" date="11/12/2007" oldid="120" color="" topic="">i will draw it..</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="185">
                <Utterance genid="186" ref="-1" time="09:49:57" date="11/12/2007" oldid="121" color="" topic="">based on some mathematical calculation, you can determine the plane</Utterance>
                <Utterance genid="187" ref="184" time="09:50:05" date="11/12/2007" oldid="122" color="" topic="">ok... , please draw it</Utterance>
                <Utterance genid="188" ref="-1" time="09:50:37" date="11/12/2007" oldid="123" color="" topic="">but, don't broke my piece of art :D</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="189">
                <Utterance genid="190" ref="184" time="09:50:53" date="11/12/2007" oldid="124" color="" topic="">:))</Utterance>
                <Utterance genid="191" ref="-1" time="09:51:01" date="11/12/2007" oldid="125" color="" topic="">you can delete them</Utterance>
                <Utterance genid="192" ref="-1" time="09:51:30" date="11/12/2007" oldid="126" color="" topic="">but the question is still in force...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="193">
                <Utterance genid="194" ref="-1" time="09:51:46" date="11/12/2007" oldid="127" color="" topic="">while you are busy with drawing I can boast a bit about where my method is used</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="195">
                <Utterance genid="196" ref="-1" time="09:51:57" date="11/12/2007" oldid="128" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="197">
                <Utterance genid="198" ref="-1" time="09:52:11" date="11/12/2007" oldid="129" color="" topic="">for instance most email clients such as Mozilla Thunderbird or Microsoft Outlook use Naive Bayes classifiers for filtering out spam emails</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="199">
                <Utterance genid="200" ref="-1" time="09:52:43" date="11/12/2007" oldid="130" color="" topic="">(#user1#) do you see my figures?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="201">
                <Utterance genid="202" ref="-1" time="09:53:04" date="11/12/2007" oldid="131" color="" topic="">nop</Utterance>
                <Utterance genid="203" ref="-1" time="09:53:14" date="11/12/2007" oldid="132" color="" topic="">i can see only mine</Utterance>
                <Utterance genid="204" ref="-1" time="09:53:44" date="11/12/2007" oldid="133" color="" topic="">let's continue a little....</Utterance>
                <Utterance genid="205" ref="-1" time="09:53:59" date="11/12/2007" oldid="134" color="" topic="">if the 2 types of data are not separable with a line</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="206">
                <Utterance genid="207" ref="-1" time="09:54:08" date="11/12/2007" oldid="135" color="" topic="">yes</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="208">
                <Utterance genid="209" ref="-1" time="09:54:08" date="11/12/2007" oldid="136" color="" topic="">(2 dimension space case)</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="210">
                <Utterance genid="211" ref="-1" time="09:54:48" date="11/12/2007" oldid="137" color="" topic="">it is my question...</Utterance>
                <Utterance genid="212" ref="-1" time="09:55:15" date="11/12/2007" oldid="138" color="" topic="">if we don't cant separate them with the line</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="213">
                <Utterance genid="214" ref="-1" time="09:55:21" date="11/12/2007" oldid="139" color="" topic="">then you can add a new dimension, so the data became separable in 3 dimension space</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="215">
                <Utterance genid="216" ref="-1" time="09:56:19" date="11/12/2007" oldid="140" color="" topic="">how many planes can accept this method??</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="217">
                <Utterance genid="218" ref="212" time="09:56:33" date="11/12/2007" oldid="141" color="" topic="">I just answer to your question</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="219">
                <Utterance genid="220" ref="-1" time="09:56:51" date="11/12/2007" oldid="142" color="" topic="">i see...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="221">
                <Utterance genid="222" ref="-1" time="09:57:06" date="11/12/2007" oldid="143" color="" topic="">the number of planes is not limited</Utterance>
                <Utterance genid="223" ref="-1" time="09:57:59" date="11/12/2007" oldid="144" color="" topic="">the only limit is that when the number of dimensions rise, to translate everything in a space with more dimension in which data is separable by a hyperplane is not a simple problem</Utterance>
                <Utterance genid="224" ref="-1" time="09:58:33" date="11/12/2007" oldid="145" color="" topic="">Because, just in 2 dimension case you add a new dimension, but in a more complex space, you need to add more than one new dimension</Utterance>
                <Utterance genid="225" ref="-1" time="09:59:34" date="11/12/2007" oldid="146" color="" topic="">And there is another problem, if you add too many dimensions, another problem appears</Utterance>
                <Utterance genid="226" ref="-1" time="10:00:20" date="11/12/2007" oldid="147" color="" topic="">you will find more than one hyperplane that separes the data into 2 classes.</Utterance>
                <Utterance genid="227" ref="-1" time="10:00:59" date="11/12/2007" oldid="148" color="" topic="">so, the number of dimmension that are added must by well chosen, based on previous experience</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="228">
                <Utterance genid="229" ref="-1" time="10:01:04" date="11/12/2007" oldid="149" color="" topic="">from the mathematic formula - the big number of planes - complicate the problem?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="230">
                <Utterance genid="231" ref="-1" time="10:01:35" date="11/12/2007" oldid="150" color="" topic="">i don't understand your question.</Utterance>
                <Utterance genid="232" ref="-1" time="10:01:38" date="11/12/2007" oldid="151" color="" topic="">I am sorry</Utterance>
                <Utterance genid="233" ref="-1" time="10:01:45" date="11/12/2007" oldid="152" color="" topic="">can you make it more specific?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="234">
                <Utterance genid="235" ref="224" time="10:01:54" date="11/12/2007" oldid="153" color="" topic="">sorry</Utterance>
                <Utterance genid="236" ref="-1" time="10:02:02" date="11/12/2007" oldid="154" color="" topic="">for my english...</Utterance>
                <Utterance genid="237" ref="-1" time="10:02:07" date="11/12/2007" oldid="155" color="" topic="">i want to ask</Utterance>
                <Utterance genid="238" ref="-1" time="10:02:49" date="11/12/2007" oldid="156" color="" topic="">if the big number of planes complicate the mathematic problem of solving ecuation...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="239">
                <Utterance genid="240" ref="-1" time="10:03:03" date="11/12/2007" oldid="157" color="" topic="">don't be shy</Utterance>
                <Utterance genid="241" ref="238" time="10:03:26" date="11/12/2007" oldid="158" color="" topic="">yeap..</Utterance>
                <Utterance genid="242" ref="241" time="10:04:01" date="11/12/2007" oldid="159" color="" topic="">when you go back into the problem numrber of dimmension</Utterance>
                <Utterance genid="243" ref="-1" time="10:04:12" date="11/12/2007" oldid="160" color="" topic="">you will have more than one solution</Utterance>
                <Utterance genid="244" ref="-1" time="10:04:35" date="11/12/2007" oldid="161" color="" topic="">if you chose too many dimension as an extension to your problem in order to classify the data.</Utterance>
                <Utterance genid="245" ref="-1" time="10:04:53" date="11/12/2007" oldid="162" color="" topic="">An interesting thing, and none of you asked me about this</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="246">
                <Utterance genid="247" ref="-1" time="10:05:11" date="11/12/2007" oldid="163" color="" topic="">so the conclusion is - to change the classifyer??:))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="248">
                <Utterance genid="249" ref="-1" time="10:05:13" date="11/12/2007" oldid="164" color="" topic="">is that until now, i talked about classifing data into 2 types</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="250">
                <Utterance genid="251" ref="249" time="10:06:00" date="11/12/2007" oldid="165" color="" topic="">so is it just a binary classifier?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="252">
                <Utterance genid="253" ref="247" time="10:06:08" date="11/12/2007" oldid="166" color="" topic="">no, the conclusion is that you must add one new dimension to the problem, until the data in that new dimmension became separable by a hyper plane</Utterance>
                <Utterance genid="254" ref="251" time="10:06:34" date="11/12/2007" oldid="167" color="" topic="">no, this is a good question, and i will give you the answer right away</Utterance>
                <Utterance genid="255" ref="-1" time="10:06:50" date="11/12/2007" oldid="168" color="" topic="">if you have to classify 3 things A,B and C</Utterance>
                <Utterance genid="256" ref="-1" time="10:07:35" date="11/12/2007" oldid="169" color="" topic="">you can proceed like this : You determine the data that is classified as A, then the data that is classified as B, and than as C</Utterance>
                <Utterance genid="257" ref="-1" time="10:07:53" date="11/12/2007" oldid="170" color="" topic="">the process needs more steps but it works in real life problems</Utterance>
                <Utterance genid="258" ref="-1" time="10:08:22" date="11/12/2007" oldid="171" color="" topic="">i don't know if i was clear about classifing data into multiple catalogs....</Utterance>
                <Utterance genid="259" ref="-1" time="10:08:25" date="11/12/2007" oldid="172" color="" topic="">?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="260">
                <Utterance genid="261" ref="-1" time="10:08:57" date="11/12/2007" oldid="173" color="" topic="">no..its ok</Utterance>
                <Utterance genid="262" ref="-1" time="10:09:01" date="11/12/2007" oldid="174" color="" topic="">go on</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="263">
                <Utterance genid="264" ref="258" time="10:09:04" date="11/12/2007" oldid="175" color="" topic="">clear enough</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="265">
                <Utterance genid="266" ref="-1" time="10:10:06" date="11/12/2007" oldid="176" color="" topic="">ok...</Utterance>
                <Utterance genid="267" ref="-1" time="10:10:23" date="11/12/2007" oldid="177" color="" topic="">i think it is (#user0#) turn to talk about his classifing method</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="268">
                <Utterance genid="269" ref="-1" time="10:10:48" date="11/12/2007" oldid="178" color="" topic="">yes</Utterance>
                <Utterance genid="270" ref="-1" time="10:10:51" date="11/12/2007" oldid="179" color="" topic="">so...</Utterance>
                <Utterance genid="271" ref="-1" time="10:11:56" date="11/12/2007" oldid="180" color="" topic="">HMM - Hidden Markov Model</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="272">
                <Utterance genid="273" ref="-1" time="10:12:15" date="11/12/2007" oldid="181" color="" topic="">yes</Utterance>
                <Utterance genid="274" ref="-1" time="10:12:29" date="11/12/2007" oldid="182" color="" topic="">can you tell us a little about it</Utterance>
                <Utterance genid="275" ref="-1" time="10:12:30" date="11/12/2007" oldid="183" color="" topic="">?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="276">
                <Utterance genid="277" ref="-1" time="10:12:39" date="11/12/2007" oldid="184" color="" topic="">is a sequence classifier</Utterance>
                <Utterance genid="278" ref="-1" time="10:13:07" date="11/12/2007" oldid="185" color="" topic="">A sequence classifier is a model whose job is to assign some label or class to each unit in a sequence</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="279">
                <Utterance genid="280" ref="-1" time="10:13:17" date="11/12/2007" oldid="186" color="" topic="">can you define a sequence classifier</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="281">
                <Utterance genid="282" ref="-1" time="10:13:25" date="11/12/2007" oldid="187" color="" topic="">:)</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="283">
                <Utterance genid="284" ref="278" time="10:13:27" date="11/12/2007" oldid="188" color="" topic="">ok..</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="285">
                <Utterance genid="286" ref="-1" time="10:13:46" date="11/12/2007" oldid="189" color="" topic="">The HMMs extend this notion by being probabilistic sequence classifiers; given a sequence of units (words, letters, morphemes, sentences, whatever) their job is to compute a probability distribution over possible labels and choose the best label sequence.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="287">
                <Utterance genid="288" ref="-1" time="10:13:53" date="11/12/2007" oldid="190" color="" topic="">cve faci ma?</Utterance>
                <Utterance genid="289" ref="-1" time="10:13:58" date="11/12/2007" oldid="191" color="" topic="">sorry</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="290">
                <Utterance genid="291" ref="-1" time="10:14:02" date="11/12/2007" oldid="192" color="" topic="">:)</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="292">
                <Utterance genid="293" ref="-1" time="10:14:02" date="11/12/2007" oldid="193" color="" topic="">about that</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="294">
                <Utterance genid="295" ref="-1" time="10:14:05" date="11/12/2007" oldid="194" color="" topic="">sorry</Utterance>
                <Utterance genid="296" ref="-1" time="10:14:09" date="11/12/2007" oldid="195" color="" topic="">do it</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="297">
                <Utterance genid="298" ref="-1" time="10:15:13" date="11/12/2007" oldid="196" color="" topic="">ok</Utterance>
                <Utterance genid="299" ref="-1" time="10:15:14" date="11/12/2007" oldid="197" color="" topic="">done</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="300">
                <Utterance genid="301" ref="-1" time="10:15:41" date="11/12/2007" oldid="198" color="" topic="">The Hidden Markov Model is one of the most important machine learning models inspeech and language processing. In order to define it properly, we need to first introducethe Markov chain, sometimes called the observed Markov model.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="302">
                <Utterance genid="303" ref="-1" time="10:15:42" date="11/12/2007" oldid="199" color="" topic="">i noticed that this "chat software" is built by Fraunhofer Institute</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="304">
                <Utterance genid="305" ref="-1" time="10:16:03" date="11/12/2007" oldid="200" color="" topic="">you somethinbg about it?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="306">
                <Utterance genid="307" ref="-1" time="10:16:08" date="11/12/2007" oldid="201" color="" topic="">i feel that i must inform you, that i done my diploma thesis there:d</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="308">
                <Utterance genid="309" ref="-1" time="10:16:16" date="11/12/2007" oldid="202" color="" topic="">WOW</Utterance>
                <Utterance genid="310" ref="-1" time="10:16:17" date="11/12/2007" oldid="203" color="" topic="">:))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="311">
                <Utterance genid="312" ref="-1" time="10:16:33" date="11/12/2007" oldid="204" color="" topic="">ok...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="313">
                <Utterance genid="314" ref="307" time="10:16:44" date="11/12/2007" oldid="205" color="" topic="">very nice</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="315">
                <Utterance genid="316" ref="-1" time="10:16:52" date="11/12/2007" oldid="206" color="" topic="">so it is used mostly in speach proccessing</Utterance>
                <Utterance genid="317" ref="316" time="10:17:08" date="11/12/2007" oldid="207" color="" topic="">processing</Utterance>
                <Utterance genid="318" ref="-1" time="10:18:30" date="11/12/2007" oldid="208" color="" topic=""/>
            </Turn>
            <Turn nickname="(#user0#)" genid="319">
                <Utterance genid="320" ref="-1" time="10:19:08" date="11/12/2007" oldid="209" color="" topic="">I import the figure that shows the general architecture of an instantiated HMM...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="321">
                <Utterance genid="322" ref="-1" time="10:19:16" date="11/12/2007" oldid="210" color="" topic="">great</Utterance>
                <Utterance genid="323" ref="-1" time="10:19:29" date="11/12/2007" oldid="211" color="" topic="">this way should be easier to understand</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="324">
                <Utterance genid="325" ref="-1" time="10:19:45" date="11/12/2007" oldid="212" color="" topic="">yes...</Utterance>
                <Utterance genid="326" ref="-1" time="10:20:08" date="11/12/2007" oldid="213" color="" topic="">Each oval shape represents a random variable that can adopt a number of values. The random variable x(t) is the hidden state at time t ...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="327">
                <Utterance genid="328" ref="-1" time="10:21:53" date="11/12/2007" oldid="214" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="329">
                <Utterance genid="330" ref="-1" time="10:22:00" date="11/12/2007" oldid="215" color="" topic="">(with the model from the above diagram)</Utterance>
                <Utterance genid="331" ref="-1" time="10:22:50" date="11/12/2007" oldid="216" color="" topic="">The random variable y(t) is the observation at time t...</Utterance>
                <Utterance genid="332" ref="-1" time="10:23:09" date="11/12/2007" oldid="217" color="" topic="">The arrows in the diagram (often called a trellis diagram) denote conditional dependencies.</Utterance>
                <Utterance genid="333" ref="-1" time="10:23:20" date="11/12/2007" oldid="218" color="" topic="">From the diagram, it is clear that the value of the hidden variable x(t) (at time t) only depends on the value of the hidden variable x(t - 1) (at time t - 1). This is called the Markov property. Similarly, the value of the observed variable y(t) only depends on the value of the hidden variable x(t) (both at time t).</Utterance>
                <Utterance genid="334" ref="-1" time="10:24:30" date="11/12/2007" oldid="219" color="" topic="">There are two strong reasons why this has occurred. First the models are very rich in mathematical structure and hence can form the theoretical basis for use in a wide range of applications. Second the models, when applied properly, work very well in practice for several important applications. In this paper we attempt to carefully and methodically review the theoretical aspects of this type of statistical modeling and show how they have been applied to selected problems in machine recognition of speech.</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="335">
                <Utterance genid="336" ref="334" time="10:25:34" date="11/12/2007" oldid="220" color="" topic="">pl</Utterance>
                <Utterance genid="337" ref="-1" time="10:25:38" date="11/12/2007" oldid="221" color="" topic="">ok i mean</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="338">
                <Utterance genid="339" ref="334" time="10:26:42" date="11/12/2007" oldid="222" color="" topic="">Hidden Markov models are especially known for their application in temporal pattern recognition such as speech, handwriting, gesture recognition, musical score following, partial discharges and bioinformatics...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="340">
                <Utterance genid="341" ref="-1" time="10:26:47" date="11/12/2007" oldid="223" color="" topic="">i think the subject you will present will be very interesting</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="342">
                <Utterance genid="343" ref="-1" time="10:26:54" date="11/12/2007" oldid="224" color="" topic="">yes</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="344">
                <Utterance genid="345" ref="-1" time="10:27:00" date="11/12/2007" oldid="225" color="" topic="">ok, go ahead</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="346">
                <Utterance genid="347" ref="-1" time="10:28:01" date="11/12/2007" oldid="226" color="" topic="">i thing its time to decide the best classifyer...</Utterance>
                <Utterance genid="348" ref="-1" time="10:28:12" date="11/12/2007" oldid="227" color="" topic="">for our needs...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="349">
                <Utterance genid="350" ref="-1" time="10:28:22" date="11/12/2007" oldid="228" color="" topic="">i need to add something to my presentation : SVMs are very flexible even because of the so called kernel functions, that map the input data into o space with more dimensions, but where the data can be separated by a hyperplane more easily</Utterance>
                <Utterance genid="351" ref="-1" time="10:29:04" date="11/12/2007" oldid="229" color="" topic="">Ok, let's first formulate the problem</Utterance>
                <Utterance genid="352" ref="-1" time="10:29:53" date="11/12/2007" oldid="230" color="" topic="">we need a classifier that would be the best for speach recognition, isn't it?</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="353">
                <Utterance genid="354" ref="-1" time="10:30:06" date="11/12/2007" oldid="231" color="" topic="">yes</Utterance>
                <Utterance genid="355" ref="-1" time="10:30:14" date="11/12/2007" oldid="232" color="" topic="">in this case - HMM is the best</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="356">
                <Utterance genid="357" ref="-1" time="10:31:06" date="11/12/2007" oldid="233" color="" topic="">this is true from my point of view as well</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="358">
                <Utterance genid="359" ref="-1" time="10:32:08" date="11/12/2007" oldid="234" color="" topic="">so give our facts...</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="360">
                <Utterance genid="361" ref="-1" time="10:32:16" date="11/12/2007" oldid="235" color="" topic="">NB is good in for rather simple tasks when we need one to be computationally fast</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="362">
                <Utterance genid="363" ref="-1" time="10:32:32" date="11/12/2007" oldid="236" color="" topic="">and i give you my...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="364">
                <Utterance genid="365" ref="-1" time="10:32:51" date="11/12/2007" oldid="237" color="" topic="">(HMMs) are, undoubtedly, the most employed coretechnique for Automatic Speech Recognition (ASR)</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="366">
                <Utterance genid="367" ref="-1" time="10:33:08" date="11/12/2007" oldid="238" color="" topic="">YES</Utterance>
                <Utterance genid="368" ref="-1" time="10:33:10" date="11/12/2007" oldid="239" color="" topic="">:))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="369">
                <Utterance genid="370" ref="-1" time="10:33:12" date="11/12/2007" oldid="240" color="" topic="">but, The Support Vector Machines (SVMs) are effective discriminant classifiers capable ofmaximizing the error margin.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="371">
                <Utterance genid="372" ref="-1" time="10:33:22" date="11/12/2007" oldid="241" color="" topic="">:P</Utterance>
                <Utterance genid="373" ref="-1" time="10:33:26" date="11/12/2007" oldid="242" color="" topic="">may be</Utterance>
                <Utterance genid="374" ref="-1" time="10:33:27" date="11/12/2007" oldid="243" color="" topic="">but</Utterance>
                <Utterance genid="375" ref="-1" time="10:33:34" date="11/12/2007" oldid="244" color="" topic="">A Markov chain is useful when we need to compute a probability for a sequence ofevents that we can observe in the world.</Utterance>
                <Utterance genid="376" ref="-1" time="10:33:44" date="11/12/2007" oldid="245" color="" topic="">In many cases, however, the events we areinterested in may not be directly observable in the world. For example, in part-of-speechtagging we didn't observe part of speech tags in the world; we sawwords, and had to infer the correct tags from the word sequence.</Utterance>
                <Utterance genid="377" ref="-1" time="10:33:54" date="11/12/2007" oldid="246" color="" topic="">We call the part-of-speech tags hidden because they are not observed. The same architecture will comeup in speech recognition; in that case we'll see acoustic events in the world, and haveto infer the presence of 'hidden' words that are the underlying causal source of theacoustics. A Hidden Markov Model (HMM) allows us to talk about both observed events (like words that we see in the input) and hidden events (like part-of-speech tags)that we think of as causal factors in our probabilistic model.</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="378">
                <Utterance genid="379" ref="-1" time="10:34:11" date="11/12/2007" oldid="247" color="" topic="">I believe if we combine our methods in a single aplication we will obtain good results</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="380">
                <Utterance genid="381" ref="-1" time="10:34:12" date="11/12/2007" oldid="248" color="" topic="">waw, calm down:d</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="382">
                <Utterance genid="383" ref="-1" time="10:34:19" date="11/12/2007" oldid="249" color="" topic="">:))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="384">
                <Utterance genid="385" ref="-1" time="10:35:06" date="11/12/2007" oldid="250" color="" topic="">the problem with SVMs is the lack of ability to handle vectors of different dimensions... and in real life, words have different number of letters</Utterance>
                <Utterance genid="386" ref="-1" time="10:35:42" date="11/12/2007" oldid="251" color="" topic="">With a modified version of SVMs capable to handle variable vector dimmensions</Utterance>
                <Utterance genid="387" ref="-1" time="10:36:01" date="11/12/2007" oldid="252" color="" topic="">the problem of speech recognition it is not more a problem</Utterance>
                <Utterance genid="388" ref="-1" time="10:36:22" date="11/12/2007" oldid="253" color="" topic="">another limitationis that SVMs only classify, but they don't give us a reliable measure of the probabilityof the correctness of the classification</Utterance>
                <Utterance genid="389" ref="-1" time="10:37:02" date="11/12/2007" oldid="254" color="" topic="">There are some known variantes of SVMs which can handle speach recognition, but i must admit that HMM is better</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="390">
                <Utterance genid="391" ref="379" time="10:37:33" date="11/12/2007" oldid="255" color="" topic="">as it has been done in text categorization for multi-page documents. A Hybrid Naive Bayes HMM ApproachEmpirical evaluation indicates that the error rate can be roughly reduced by half contextual information is incorporated.</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="392">
                <Utterance genid="393" ref="-1" time="10:38:21" date="11/12/2007" oldid="256" color="" topic="">ok</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="394">
                <Utterance genid="395" ref="-1" time="10:38:41" date="11/12/2007" oldid="257" color="" topic="">every method can overperform the others, it depends on the task, the problem which needs to be solved</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="396">
                <Utterance genid="397" ref="-1" time="10:38:49" date="11/12/2007" oldid="258" color="" topic="">i think we can combine all 3 for the best results</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="398">
                <Utterance genid="399" ref="-1" time="10:38:52" date="11/12/2007" oldid="259" color="" topic="">yeap..</Utterance>
                <Utterance genid="400" ref="-1" time="10:39:04" date="11/12/2007" oldid="260" color="" topic="">SVMs has best results for handling errors</Utterance>
                <Utterance genid="401" ref="-1" time="10:39:39" date="11/12/2007" oldid="261" color="" topic="">and if first use HMM on speech recognition and then test with SVMs the results</Utterance>
                <Utterance genid="402" ref="-1" time="10:39:57" date="11/12/2007" oldid="262" color="" topic="">i think we will achieve the best results and we can help each other....</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="403">
                <Utterance genid="404" ref="-1" time="10:40:20" date="11/12/2007" oldid="263" color="" topic="">So...we must make an aliance of our firms...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="405">
                <Utterance genid="406" ref="-1" time="10:40:30" date="11/12/2007" oldid="264" color="" topic="">we can combine our companies into one that would provide the best classifiers in the world</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="407">
                <Utterance genid="408" ref="-1" time="10:40:36" date="11/12/2007" oldid="265" color="" topic="">to sign the contract..;))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="409">
                <Utterance genid="410" ref="404" time="10:40:50" date="11/12/2007" oldid="266" color="" topic="">yes, i am very happy that you understand the problem</Utterance>
                <Utterance genid="411" ref="-1" time="10:41:19" date="11/12/2007" oldid="267" color="" topic="">and the consequences if we continue to work not as partners...</Utterance>
                <Utterance genid="412" ref="-1" time="10:42:13" date="11/12/2007" oldid="268" color="" topic="">it is possible that due to tough concurrency, each of our companies to get bankrupted</Utterance>
                <Utterance genid="413" ref="-1" time="10:42:19" date="11/12/2007" oldid="269" color="" topic="">and we don't want this</Utterance>
                <Utterance genid="414" ref="-1" time="10:42:39" date="11/12/2007" oldid="270" color="" topic="">we can maximize our profits and have the best product on the market</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="415">
                <Utterance genid="416" ref="-1" time="10:43:14" date="11/12/2007" oldid="271" color="" topic="">i'm agree</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="417">
                <Utterance genid="418" ref="-1" time="10:43:25" date="11/12/2007" oldid="272" color="" topic="">that's a deal then!</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="419">
                <Utterance genid="420" ref="-1" time="10:43:36" date="11/12/2007" oldid="273" color="" topic="">If we use all these 3 classifiers, we can deliver a product that would be the best not only for speach recognition but in many industrial fields...</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="421">
                <Utterance genid="422" ref="-1" time="10:43:56" date="11/12/2007" oldid="274" color="" topic="">it's a good idea</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="423">
                <Utterance genid="424" ref="-1" time="10:44:08" date="11/12/2007" oldid="275" color="" topic="">In the begining of my presentation i told you about the success of SVMs in medicine</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="425">
                <Utterance genid="426" ref="-1" time="10:44:41" date="11/12/2007" oldid="276" color="" topic="">yes...i think the SVM is the best method in medical field...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="427">
                <Utterance genid="428" ref="-1" time="10:44:53" date="11/12/2007" oldid="277" color="" topic="">it was used to diagnosticate some cancer diseases</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="429">
                <Utterance genid="430" ref="-1" time="10:45:01" date="11/12/2007" oldid="278" color="" topic="">and I told you about e-mail clients which use NB for filtering e-mail</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="431">
                <Utterance genid="432" ref="-1" time="10:45:12" date="11/12/2007" oldid="279" color="" topic="">yes</Utterance>
                <Utterance genid="433" ref="-1" time="10:45:32" date="11/12/2007" oldid="280" color="" topic="">I know that Naive Bayes is a model based classifier</Utterance>
                <Utterance genid="434" ref="-1" time="10:47:12" date="11/12/2007" oldid="281" color="" topic="">so... with some Gaussian distribution... NB could perform best...</Utterance>
                <Utterance genid="435" ref="-1" time="10:47:45" date="11/12/2007" oldid="282" color="" topic="">I don't need to remind you that it is supposed that the world reflects the Gaussion distribution....</Utterance>
                <Utterance genid="436" ref="435" time="10:47:57" date="11/12/2007" oldid="283" color="" topic="">in many fields</Utterance>
                <Utterance genid="437" ref="-1" time="10:48:16" date="11/12/2007" oldid="284" color="" topic="">I think it is ok for this point of our business</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="438">
                <Utterance genid="439" ref="-1" time="10:48:29" date="11/12/2007" oldid="285" color="" topic="">so, should we shake hands, gentlemen?</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="440">
                <Utterance genid="441" ref="-1" time="10:48:35" date="11/12/2007" oldid="286" color="" topic="">yes</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="442">
                <Utterance genid="443" ref="-1" time="10:48:38" date="11/12/2007" oldid="287" color="" topic="">yes</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="444">
                <Utterance genid="445" ref="-1" time="10:48:47" date="11/12/2007" oldid="288" color="" topic="">and don't forget that enclosed to this meeting</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="446">
                <Utterance genid="447" ref="-1" time="10:48:49" date="11/12/2007" oldid="289" color="" topic="">i want to sleep...:((</Utterance>
                <Utterance genid="448" ref="-1" time="10:48:51" date="11/12/2007" oldid="290" color="" topic="">:))</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="449">
                <Utterance genid="450" ref="-1" time="10:49:36" date="11/12/2007" oldid="291" color="" topic="">we have some pictures with the characteristics of each classifier</Utterance>
                <Utterance genid="451" ref="-1" time="10:49:43" date="11/12/2007" oldid="292" color="" topic="">ok.. have a nice evening</Utterance>
                <Utterance genid="452" ref="-1" time="10:49:51" date="11/12/2007" oldid="293" color="" topic="">and i look forward to meet you again</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="453">
                <Utterance genid="454" ref="-1" time="10:50:21" date="11/12/2007" oldid="294" color="" topic="">yes...to speak about details of our cooperation</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="455">
                <Utterance genid="456" ref="452" time="10:50:25" date="11/12/2007" oldid="295" color="" topic="">so do I, we need to dicuss the details</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="457">
                <Utterance genid="458" ref="-1" time="10:50:31" date="11/12/2007" oldid="296" color="" topic="">if you have any misunderstandings or question, please don't hesitate to email or call me</Utterance>
            </Turn>
            <Turn nickname="(#user0#)" genid="459">
                <Utterance genid="460" ref="-1" time="10:50:55" date="11/12/2007" oldid="297" color="" topic="">so...bye bye and have a good night...</Utterance>
            </Turn>
            <Turn nickname="(#user1#)" genid="461">
                <Utterance genid="462" ref="-1" time="10:51:01" date="11/12/2007" oldid="298" color="" topic="">Regards, gentlemens</Utterance>
            </Turn>
            <Turn nickname="(#user2#)" genid="464">
                <Utterance genid="465" ref="-1" time="10:51:12" date="11/12/2007" oldid="300" color="" topic="">bye bye</Utterance>
            </Turn>
        </Body>
    </Dialog>
