Marys Medicine

Analyzing the Effect of Query Class on Document
Pawel Kowalczyk, Ingrid Zukerman, and Michael Niemann School of Computer Science and Software Engineering Monash University Clayton, VICTORIA 3800, AUSTRALIA Abstract. Analysis of queries posed to open-domain question-answering sys-
tems indicates that particular types of queries are dominant, e.g., queries about
the identity of people, and about the location or time of events. We applied a rule-
based mechanism and performed manual classification to classify queries into
such commonly occurring types. We then experimented with different adjust-
ments to our basic document retrieval process for each query type. The applica-
tion of the best retrieval adjustment for each query type yielded improvements in
retrieval performance. Finally, we applied a machine learning technique to auto-
matically learn the manually classified query types, and applied the best retrieval
adjustments obtained for the manual classification to the automatically learned
query classes. The learning algorithm exhibited high accuracy, and the retrieval
performance obtained for the learned classes was consistent with the performance
obtained for the rule-based and manual classifications.
The growth in popularity of the Internet highlights the importance of developing sys-tems that generate responses to queries targeted at large unstructured corpora. Thesequeries vary in their informational goal and topic, ranging from requests for descrip-tions of people or things, to queries about the location or time of events, and questionsabout specific attributes of people or things. There is also some variation in the successof question-answering systems in answering the different types of queries.
Recently, there has been some work on predicting whether queries can be answered by the documents in a particular corpus [1, 2]. The hope is that by identifying featuresthat affect the "answerability" of queries, the queries can be modified prior to attemptingdocument retrieval, or appropriate steps can be taken during the retrieval process toaddress problems that arise due to these features.
In this paper, we investigate the use of query type as a predictor of document re- trieval performance in the context of a question answering task, and as a basis for theautomatic selection of a retrieval policy. The first step in our study consisted of per-forming two types of query classification: a coarse-grained classification, which wasperformed by means of a rule-based mechanism, and a finer-grained classification,which was done manually. We considered these two types of classification becausefiner-grained classes are believed to be more informative than coarser-grained classes when extracting answers to queries from retrieved documents [3]. However, prior tocommitting to a particular classification grain, we must examine its effect on documentretrieval performance (as document retrieval is the step that precedes question answer-ing).
Our analysis of the effect of query type on document retrieval performance shows that performance varies across different types of queries. This led us to experiment withdifferent types of adjustments to our basic document retrieval process, in order to deter-mine the best adjustment for each type of query. The application of specific adjustmentsto the retrieval of documents for different query types yielded improvements in retrievalperformance both for the coarse-grained and the finer-grained query classes.
In the last step of our study, we applied a supervised machine learning technique, namely Support Vector Machines (SVMs) [4], to learn the finer-grained query typesfrom shallow linguistic features of queries, and used the query-type-based retrievaladjustments to retrieve documents for the learned classes.1 Our results for both themachine learning algorithm and the document retrieval process are encouraging. Thelearning algorithm exhibited high accuracy, and the resultant retrieval performance wasconsistent with the performance obtained for the rule-based and manually-derived querycategories.
In the next section, we review related research. In Section 3, we describe our doc- ument retrieval procedure. Next, we describe our data set, discuss our rule-based clas-sification process and our manual classification process, and present the results of ourexperiments with the adjustments to the retrieval procedure. In Section 5, we describethe data used to train the SVM for query classification, and evaluate the performanceof the SVM and the retrieval performance for the automatically learned classes. In Sec-tion 6, we summarize the contribution of this work.
Our research is at the intersection of query classification systems [5, 6, 3] and perfor-mance prediction systems [1, 2].
Query classification systems constitute a relatively recent development in Informa- tion Retrieval (IR). Radev et al. [5] and Zhang and Lee [3] studied automatic queryclassification based on the type of the expected answer. Their work was motivated bythe idea that such a classification can help select a suitable answer in a document whenperforming an open-domain question-answering task. Radev et al. compared a machinelearning approach with a heuristic (hand-engineered) approach for query classification,and found that the latter approach outperformed the former. Zhang and Lee experi-mented with five machine learning methods to learn query classes, and concluded thatwhen only surface text features are used, SVMs outperform the other techniques. Kangand Kim's study of query classification [6] was directed at categorizing queries accord-ing to the task at hand (informational, navigational or transactional). They postulatedthat appropriate query classification supports the application of algorithms dedicatedto particular tasks. Our work resembles that of Zhang and Lee in its use of SVMs.
1 The automation of the finer-grained classification is necessary in order to incorporate it as a step in an automatic document-retrieval and question-answering process.
However, they considered two grains of classifications: coarse (6 classes) and fine (50classes), while we consider an intermediate grain (11 classes). More importantly, likeKang and Kim, we adjust our retrieval policy based on query class. However, Kang andKim's classes were broad task-oriented classes, while we offer finer distinctions withinthe informational task.
Performance-prediction systems identify query features that predict retrieval per- formance. The idea is that queries that appear "unpromising" can be modified prior toattempting retrieval, or retrieval behaviour can be adjusted for such queries. Cronen-Townsend et al. [1] developed a clarity score that measures the coherence of the lan-guage used in documents which "generate" the terms in a query. Thus, queries with ahigh clarity score yield a cohesive set of documents, while queries with a low clarityscore yield documents about different topics. Zukerman et al. [2] adopted a machinelearning approach to predict retrieval performance from the surface features of queriesand word frequency counts in the corpus. They found that queries were "answerable"when they did not contain words whose frequency exceeded a particular threshold (thisthreshold is substantially lower than the frequency of stop words, which are normallyexcluded from the retrieval process). This finding led to the automatic removal of suchwords from queries prior to document retrieval, yielding significant improvements inretrieval performance. The work described in this paper predicts retrieval performancefrom surface features of queries (by first using these features to classify queries). How-ever, it does not consider corpus-related information. Additionally, unlike the systemdescribed in Zukerman et al. that modifies the queries, our system dynamically adjustsits retrieval behaviour.2 Our retrieval mechanism combines the classic vector-space model [7] with a paraphrase-based query expansion process [8, 2]. This mechanism is further adjusted by consider-ing different numbers of paraphrases (between 0 and 19) and different retrieval policies.
Below we describe our basic retrieval procedure followed by the adjustments.
1. Tokenize, tag and lemmatize the query. Tagging is performed using Brill's part-of-speech tagger [9]. Lemmatizing consistsof converting words into lemmas, which are uninflected versions of words.
2. Generate replacement lemmas for each content lemma in the query. The replacement lemmas are the intersection of lemmas obtained from two re-sources: WordNet [10] and a thesaurus that was automatically constructed fromthe Oxford English Dictionary. The thesaurus also yields similarity scores betweeneach query lemma and its replacement lemmas.
3. Propose paraphrases for the query using different combinations of replacement lemmas, compute the similarity score between each paraphrase and the query, and 2 We also replicated Zukerman et al.'s machine learning experiments. However, since our docu- ment retrieval technique combines the vector-space model with boolean retrieval (Section 3),the results obtained when their machine learning approach was used with our system differedfrom Zukerman et al.'s original findings.
rank the paraphrases according to their score.
The similarity score of each paraphrase is computed from the similarity scores be-tween the original lemmas in the query and the corresponding replacement lemmasin the paraphrase.
4. Retain the lemmatized query plus the top K paraphrases (the default value for K 5. Retrieve documents for the query and its paraphrases using a paraphrase-adjusted version of the vector-space model.
For each lemma in the original query or its paraphrases, documents that containthis lemma are retrieved. Each document is scored using a function that combinesthe tf.idf (term frequency inverse document frequency) score [7] of the query lem-mas and paraphrase lemmas that appear in the document, and the similarity scorebetween the paraphrase lemmas that appear in the document and the query lemmas.
The tf.idf part of the score takes into account statistical features of the corpus, andthe similarity part takes into account semantic features of the query.
6. Retain the top N documents (at present N = 200).
Adjustments to the basic retrieval procedure
The adjustments to our retrieval procedure pertain to the number of paraphrases usedfor query expansion and to the retrieval policy used in combination with the vectorspace model. The effect of these adjustments on retrieval performance is discussed inSections 4.1 and 4.2.
Number of paraphrases. We consider different numbers of paraphrases (between 0 and19), in addition to the original query.
Retrieval policies. Our system features three boolean document retrieval policies, whichare used to constrain the output of the vector-space model: (1) 1NNP, (2) 1NG (1 NounGroup), and (3) MultipleNGs.
1NNP – retrieve documents that contain at least one proper noun (NNP) from the query.3 If no proper nouns are found, fall back to the vector space model.
1NG – retrieve documents that contain the content words of at least one of the noun groups in the query, where a noun group is a sequence of nouns possibly interleavedby adjectives and function words, e.g., "Secretary of State", "pitcher's mound" or"house".
MultipleNGs – retrieve documents that contain at least g NGs in the query, where g = min{2, # of NGs in the query}.
3 NNP is the tag used for singular proper nouns in parsers and part-of-speech taggers. This tag is part of the Penn Treebank tag-set (
Query Classification and Retrieval Adjustment
Our dataset consists of 911 unique queries from the TREC11 and TREC12 corpora.
These queries were obtained from logs of public repositories such MSNSearch andAskJeeves. Their average length is 8.9 words, with most queries containing between 5and 12 words. The answers to these queries are retrieved from approximately 1 milliondocuments in the ACQUAINT corpus (this corpus is part of the NIST Text ResearchCollection, These documents are newspaper articles fromthe New York Times, Associated Press Worldstream (APW), and Xinhua English (Peo-ple's Republic of China) news services. Thus, the task at hand is an example of the moregeneral problem of finding answers to questions in open-domain documents that werenot designed with these questions in mind (in contrast to encyclopedias).
We first extracted six main query features by automatically performing shallow lin- guistic analysis of the queries. These features are 1. Type of the initial query words – corresponds mostly to the first word in the query, but merges some words, such as "what" and "which", into a single category, andconsiders additional words if the first word is "how".
2. Main focus – the attribute sought in the answer to the query, e.g., "How far is it from Earth to Mars?" (similar components have been considered in [11, 12]).
3. Main verb – the main content verb of the query (different from auxiliary verbs such as "be" and "have"), e.g., "What book did Rachel Carson write in 1962?". It oftencorresponds to the head verb of the query, but it may also be a verb embeddedin a subordinate clause, e.g., "What is the name of the volcano that destroyed theancient city of Pompeii?".
4. Rest of the query – e.g., "What is the name of the volcano that destroyed the ancient city of Pompeii?".
5. Named entities – entities characterized by sequences of proper nouns, possibly in- terleaved with function words, e.g., "Hong Kong" or "Hunchback of Notre Dame".
6. Prepositional phrases – e.g., "In the bible, who is Jacob's mother?".
After the shallow analysis, the queries were automatically classified by a rule-based system into six broad categories which represent the type of the desired answer.
1. location, e.g., "In what country did the game of croquet originate?".
2. name, e.g., "What was Andrew Jackson's wife's name?".
3. number, e.g., "How many chromosomes does a human zygote have?".
4. person, e.g., "Who is Tom Cruise married to?".
5. time, e.g., "What year was Alaska purchased?".
6. other, which is the default category, e.g., "What lays blue eggs?".
The rules for query classification considered two main factors: type of the initial query words (feature #1 above), and main-focus words (feature #2). The first factorwas used to identify location, number, person and time queries ("where", "how [much many ADJ]", "[who whom whose]" and "when" respectively). The main focuswords were then used to classify queries whose initial word is "what", "which" or "list".
For example, "country", "state" and "river" indicate location (e.g., "What is the state Query type # of queries # of queries
with answers
ans queries (%) #para/policy ans queries (%)
Table 1. Breakdown of automatically derived categories for TREC11 and TREC12 queries; per-
formance for 19 paraphrases and 1NNP retrieval policy; best retrieval adjustment and best per-
formance (measured in answerable queries)
with the smallest population?"), and "date" and "year" indicate time (e.g., "What yearwas the light bulb invented?").
Table 1 shows the breakdown of the six query categories, together with the retrieval performance when using our default retrieval method (19 paraphrases and the 1NNPretrieval policy), and when using the retrieval adjustment that yields the best retrievalperformance for each query class. The first column lists the query type, the second col-umn shows the number of queries of this type, and the third column shows the numberof queries for which TREC participants found answers in the corpus (obtained from theTREC judgment file).4 The retrieval performance for our default method appears in thefourth column. The fifth and sixth columns present information pertaining to the bestperformance (discussed later in this section).
We employ a measure called number of answerable queries to assess retrieval per- formance. This measure, which was introduced in [2], returns the number of queriesfor which the system has retrieved at least one document that contains the answer to aquery. We use this measure because the traditional precision measure is not sufficientlyinformative in the context of a question-answering task. For instance, consider a sit-uation where 10 correct documents are retrieved for each of 2 queries and 0 correctdocuments for each of 3 queries, compared to a situation where 2 correct documentsare retrieved for each of 5 queries. Average precision would yield a better score for thefirst situation, failing to address the question of interest for the question-answering task,namely how many queries have a chance of being answered, which is 2 in the first caseand 5 in the second case.
As seen from the results in the first four columns of Table 1, there are differences in retrieval performance for the six categories. Also, the other category (which is ratheruninformative) dominates, and exhibits the worst retrieval performance. This led to twodirections of investigation: (1) examine the effect of number of paraphrases and retrieval 4 TREC releases a judgment file for all the answers submitted by participants. For each submit- ted answer (and source document for that answer) the file contains a number which representsa degree of correctness. Thus, at present, when assessing the performance of our retrieval pro-cedure, we are bounded by the answers found by previous TREC participants.
policy on retrieval performance, and (2) refine the query classification to increase thespecificity of the "other" category in particular.
Effect of number of paraphrases and retrieval policy on performance
We ran all the combinations of number of paraphrases and retrieval policies on our sixquery classes (a total of 60 runs: 20 × 3), and selected the combination of number ofparaphrases and retrieval policy that gave the best result for each query type (where sev-eral adjustments yielded the same performance, the adjustment with the lowest numberof paraphrases was selected). The fifth and sixth columns in Table 1 show the results ofthis process. The fifth column shows the number of paraphrases and retrieval policy thatyield the best performance for a particular query type, and the sixth column shows thenumber of answerable queries obtained by these adjustments. As can be seen from theseresults, the retrieval adjustments based on query type yield substantial performance im-provements.
Manual refinement of query classes
We re-examined the automatically derived query classes with a view towards a moreprecise identification of the type of the desired answer (as stated above, the hope isthat this more precise categorization will help to find answers in documents). At thesame time, we endeavoured to define categories that had some chance of being au-tomatically identified. This led us to the 11 categories specified below. These classesinclude five of the six previously defined categories (name was split between personand attribute). The queries in the six original classes were then manually re-taggedwith the new classes.
1. location, e.g., "In what country did the game of croquet originate?".
2. number, e.g., "How many chromosomes does a human zygote have?".
3. person, e.g., "Who is Tom Cruise married to?".
4. time, e.g., "What year was Alaska purchased?".
5. attribute – an attribute of the query's topic, e.g., "What is Australia's national
6. howDoYouSay – the spelling-out of an acronym or the translation of a word, e.g.,
"What does DNA stand for?".
7. object – an object or the composition of an object, e.g., "What did Alfred Nobel
8. organization – an organization or group of people, e.g., "What company manufac-
tures Sinemet?".
9. process – a process or how an event happened, e.g., "How did Mahatma Gandhi
die?". It is worth noting that 80% of the queries in this category are about howsomebody died.
10. term – a word that defines a concept, e.g., "What is the fear of lightning called?".
11. other – queries that did not fit in the other categories, e.g., "What is the chemical
formula for sulfur dioxide?".
Query type # of queries # of queries
with answers
ans queries (%) #para/policy ans queries (%)
Table 2. Breakdown of manually tagged categories for TREC11 and TREC12 queries; perfor-
mance for 19 paraphrases and 1NNP retrieval policy; best retrieval adjustment and best perfor-
mance (measured in answerable queries)
Table 2 shows the breakdown of the 11 query categories (the original categories are asterisked), together with the retrieval performance obtained for our default retrievalpolicy (19 paraphrases, 1NNP). As for Table 1, the first column lists the query types, thesecond column shows the number of queries of each type, and the third column showsthe number of queries which were deemed correct according to the TREC judgment file.
The retrieval performance for our default method appears in the fourth column. The fifthand sixth columns contain the retrieval adjustments yielding the best performance, andthe result obtained by these adjustments, respectively. As can be seen from these results,the retrieval adjustments based on the finer-grained, manually-derived query types yieldperformance improvements that are similar to those obtained for the coarser rule-basedquery types.
It is worth noting that there is nothing intrinsically important that distinguishes these 11 categories from other options. The main factor is their ability to improve systemperformance, which spans two aspects of the system: document retrieval and answerextraction. Since the finer categories are more informative than the coarser ones, thehope is that they will assist during the answer extraction stage. Our results show thatthis will not occur at the expense of retrieval performance.
Using Support Vector Machines to Learn Query Classes
The SVM representation of each query has 11 parts, which may be roughly dividedinto three sections: coarse properties (3 parts), fine-grained properties (6 parts), andWordNet properties (2 parts).
Coarse properties – these are properties that describe a query in broad terms.
• headTarget – the target or topic of the question, which is the first sequence of proper nouns in the query, and if no proper nouns are found then it is the first noun group, e.g., for the query "Who is Tom Cruise married to?", theheadTarget is "Tom Cruise".
• headConcept – the attribute we want to find out about the target, e.g., for the query "What is the currency of China?", the headConcept is "currency".
• headAction – the action performed on the target, which is mostly the head verb of the query, e.g., "married" in the above query about Tom Cruise.
Fine properties – these properties correspond to the six query features extracted
from the query by performing shallow linguistic analysis (Section 4): (1) type of theinitial query words, (2) main focus, (3) main verb, (4) rest of the query, (5) namedentities, and (6) prepositional phrases. They provide additional detail about a queryto that provided by the coarse properties (but main focus and main verb often over-lap with headConcept and headAction respectively).
WordNet properties – these properties contain the WordNet categories for the top
four WordNet senses of the main verb and main focus of the query.5 • verbWNcat – e.g., "marry" has two senses, both of which are social, yielding the value social: 2.
• focusWNcat – e.g., "currency" has four senses, two of which are attribute (which is different from our attribute query type), one possession, and onestate, yielding the values attribute: 2, possession: 1, state: 1.
Each of these parts contains a bag-of-lemmas (recall that words are lemmatized), which are modified as follows.
– Proper nouns are replaced by designations which represent how many consecutive proper nouns are in a query, e.g., "Who is Tom Cruise married to?" yields who be2NNP marry.
– Similarly, abbreviations are replaced by their designation.
– Certain combinations of up-to three query-initial words are merged, e.g., "what is the" and "who is".
The SVM was trained as follows. For each query type, we separated the 911 queries into two groups: queries that belong to that type and the rest of the queries. For instance,when training to identify person queries, our data consisted of 118 positive samplesand 793 negative samples. Each group was then randomly split in half, where one halfwas used for training and the other half for testing, e.g., for our person example, boththe training set and the test set consisted of 59 positive samples and 397 negative sam-ples.6 5 These properties perform word-sense collation, rather than word-sense disambiguation.
6 We also used another training method where the 911 queries were randomly split into two halves: training and testing. We then used the queries of a particular class in the training set,e.g., person, as positive samples (and the rest of the training queries as negative samples).
Similarly any queries of that class that were found in the test set were considered positivesamples (and the rest of the queries were negative samples). Although both methods yieldedconsistent results, we prefer the method described in the body of the paper, as it guarantees aconsistent number of positive training samples.
Query type # of queries # of queries
(with ans)
(average over 20 runs)
(avg. 20 runs,
Recall (STDV) Precision (STDV) ans queries (STDV)
0.93 (0.04) 87.8% 0.99 (0.01) 82.6% 0.82 (0.04) 85.0% 0.99 (0.01) 88.7% 0.90 (0.04) 74.7% 0.81 (0.09) 62.3% 0.90 (0.12) 82.7% 0.91 (0.09) 91.9% 0.99 (0.02) 85.1% 0.83 (0.04) 46.2% 0.77 (0.23) 68.2% Table 3. Recall and precision obtained by SVM for 11 manually derived categories for TREC11
and TREC12 queries; average retrieval performance for SVM-learned queries
Table 3 shows the recall and precision of the SVM for the 11 manually-derived query types, where recall and precision are defined as follows.
number of queries in class i learned by the SVM number of queries in class i number of queries in class i learned by the SVM Precision = number of queries attributed by the SVM to class i Also shown is the retrieval performance for the learned classes after the application ofquery-type-based retrieval adjustments. Both results were obtained from 20 trials.
As seen from the results in Table 3, six of the learned classes had over 93% recall, and seven classes had over 90% precision. The other class had a particularly low re-call (42%) and also a rather low precision (77%). However, this is not surprising owingto the amorphous nature of the queries in this class (i.e., they had no particular distin-guishing features, so queries belonging to other classes found their way into other,and other queries wandered to other classes). The organization class had a highprecision, but a lower recall, as some organization queries were wrongly identi-fied as person queries (this is a known problem in question answering, as it is oftenhard to distinguish between organizations and people without domain knowledge). Theobject class exhibited this problem to a lesser extent. Some attribute queries weremis-classified as location queries and others as person queries. This explains thelower recall obtained for attribute, and together with the organization problemmentioned above, also explains the lower precision obtained for person (location isless affected owing to the larger number of location queries).
Overall, although we used only a modified bag-of-words approach, we obtained bet- ter results for our fine-grained classification than those obtained by Zhang and Lee [3] for a coarse classification (the best performance they obtained with an SVM for bag-of-words was 85.8%, for n-grams 87.4%, and for a kernel that takes into account syntacticstructure 90.0%). This may be attributed to our use of WordNet properties, and our dis-tinction between coarse, fine and WordNet properties. Also, it is worth noting that thefeatures considered significant by the SVM for the different classes are intuitively ap-pealing. For instance, examples of the modified lemmas considered significant for thelocation class are: "city", "country", "where" and "where be", while examples of themodified lemmas for the time class are: "year", "date", "when" and "when be".
The retrieval performance shown in the last column of Table 3 is consistent with that shown in Table 2. That is, the retrieval performance obtained for the automatically-learned classes was not significantly different from that obtained for the manually-tagged classes.
We have studied two aspects of the question-answering process – performance predic-tion and query classification – and offered a new contribution at the intersection ofthese aspects: automatic selection of adjustments to the retrieval procedure based onquery class. Overall, our results show that retrieval performance can be improved bydynamically adjusting the retrieval process on the basis of automatically learned queryclasses.
In query classification, we applied SVMs to learn query classes from manually tagged queries. Although our input was largely word based, our results (averaged over20 runs) were superior to recent results in the literature. This may be attributed to thebreakdown of query properties into coarse-grained, fine-grained and WordNet-based.
In performance prediction, we first used coarse-grained query classes learned by a rule-based system as predictors of retrieval performance, and as a basis for the dynamicselection of adjustments to the retrieval procedure. This yielded improvements in re-trieval performance. Next, finer-grained, manually-derived classes were used as the ba-sis for the dynamic selection of retrieval adjustments. Retrieval performance was main-tained for these finer-grained categories. This is an encouraging result, as fine-grainedcategories are considered more useful than coarse-grained categories for answer extrac-tion. Finally, the retrieval adjustments were applied to the SVM-learned, fine-grainedquery categories, yielding a retrieval performance that is consistent with that obtainedfor the manually-derived categories. This result demonstrates the applicability of ourtechniques to an automated question-answering process.
This research is supported in part by the ARC Centre for Perceptive and IntelligentMachines in Complex Environments. The authors thank Oxford University Press forthe use of their electronic data, and Tony for developing the thesaurus.
1. Cronen-Townsend, S., Zhou, Y., Croft, W.B.: Predicting query performance. In: SIGIR'02 – Proceedings of the 25th ACM International Conference on Research and Development inInformation Retrieval, Tampere, Finland (2002) 299–306 2. Zukerman, I., Raskutti, B., Wen, Y.: Query expansion and query reduction in document retrieval. In: ICTAI2003 – Proceedings of the 15th International Conference on Tools withArtificial Intelligence, Sacramento, California (2003) 552–559 3. Zhang, D., Lee, W.S.: Question classification using Support Vector Machines. In: SIGIR'03 – Proceedings of the 26th ACM International Conference on Research and Development inInformation Retrieval, Toronto, Canada (2003) 26–32 4. Vapnik, V.: Statistical Learning Theory. Wiley-Interscience, New York (1998)5. Radev, D., Fan, W., Qi, H., Wu, H., Grewal, A.: Probabilistic question answering from the Web. In: WWW2002 – Proceedings of the 11th World Wide Web Conference, Honolulu,Hawaii (2002) 408–419 6. Kang, I.H., Kim, G.: Query type classification for Web document retrieval. In: SIGIR'03 – Proceedings of the 26th ACM International Conference on Research and Development inInformation Retrieval, Toronto, Canada (2003) 64–71 7. Salton, G., McGill, M.: An Introduction to Modern Information Retrieval. McGraw Hill 8. Zukerman, I., Raskutti, B.: Lexical query paraphrasing for document retrieval. In: COL- ING'02 – Proceedings of the International Conference on Computational Linguistics, Taipei,Taiwan (2002) 1177–1183 9. Brill, E.: A simple rule-based part of speech tagger. In: ANLP-92 – Proceedings of the Third Conference on Applied Natural Language Processing, Trento, IT (1992) 152–155 10. Miller, G., Beckwith, R., Fellbaum, C., Gross, D., Miller, K.: Introduction to WordNet: An on-line lexical database. Journal of Lexicography 3 (1990) 235–244
11. Moldovan, D., Harabagiu, S., Pasca, M., Mihalcea, R., Girju, R., Goodrum, R., Rus, V.: The structure and performance of an open-domain question answering system. In: ACL2000 –Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics,Hong Kong (2000) 563–570 12. Zukerman, I., Horvitz, E.: Using machine learning techniques to interpret WH-questions.
In: ACL01 Proceedings – the 39th Annual Meeting of the Association for ComputationalLinguistics, Toulouse, France (2001) 547–554


YOUR NEW MEDICINES Your doctors have started you on many new medicines. These medicines are needed to make sure your transplant is successful. The medicines can be put into three different classes:• Anti-rejection – Help prevent your immune system from recognizing your new lung as foreign • Anti-infective – Help protect you against certain infections• Miscellaneous – Medicines that treat the side effects of the

Skew Deviation Gennadiy Shevchenko, MD and Robert A Taylor, MD Department of Neurology, University of Minnesota A 31-year-old female presented to emergency department (ED) with acute onset of double vision, vertigo, balance difficulty, nausea, and emesis of one-day duration. Her symptoms were better with closed eyes, and diplopia was worse on left gaze with images one above the other. Her past medical history was significant for cerebellar astrocytoma at the age of 4 (which was treated with resection and location radiation therapy), seizure disorder, and ventriculoperitoneal shunt placement for treatment of hydrocephalus. She also had known hypertension. Her home medications included oxcarbazepine, verapamil, and oral contraceptive pills (OCP).