{ "paper_id": "M91-1024", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T03:15:24.915579Z" }, "title": "HUGHES TRAINABLE TEXT SKIMMER : DESCRIPTION OF THE TTS SYSTEM AS USED FOR MUC-3", "authors": [ { "first": "Charles", "middle": [ "P" ], "last": "Dolan", "suffix": "", "affiliation": { "laboratory": "Hughes Research Laboratorie s", "institution": "", "location": { "addrLine": "3011 Malibu Canyon Road M/S RL96 Malibu", "postCode": "90265", "region": "CA" } }, "email": "" }, { "first": "Seth", "middle": [ "R" ], "last": "Goldman", "suffix": "", "affiliation": { "laboratory": "Hughes Research Laboratorie s", "institution": "", "location": { "addrLine": "3011 Malibu Canyon Road M/S RL96 Malibu", "postCode": "90265", "region": "CA" } }, "email": "" }, { "first": "Thomas", "middle": [ "V" ], "last": "Cuda", "suffix": "", "affiliation": { "laboratory": "Hughes Research Laboratorie s", "institution": "", "location": { "addrLine": "3011 Malibu Canyon Road M/S RL96 Malibu", "postCode": "90265", "region": "CA" } }, "email": "" }, { "first": "Alan", "middle": [ "M" ], "last": "Nakamura", "suffix": "", "affiliation": { "laboratory": "Hughes Research Laboratorie s", "institution": "", "location": { "addrLine": "3011 Malibu Canyon Road M/S RL96 Malibu", "postCode": "90265", "region": "CA" } }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "The objective of the Hughes Trainable Text Skimmer (TTS) Project is to create text skimmin g software that: (1) can be easily re-configured for new applications, (2) improves its performanc e with use, and (3) is fast enough to process megabytes of text per day. The TTS-MUC3 system i s our first full scale prototype .", "pdf_parse": { "paper_id": "M91-1024", "_pdf_hash": "", "abstract": [ { "text": "The objective of the Hughes Trainable Text Skimmer (TTS) Project is to create text skimmin g software that: (1) can be easily re-configured for new applications, (2) improves its performanc e with use, and (3) is fast enough to process megabytes of text per day. The TTS-MUC3 system i s our first full scale prototype .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "TTS-MUC3 incorporates semi-automated lexicon generation and almost fully automated phras e pattern generation. Associative retrieval from a case memory provides raw data for computing se t fills and string fills . TTS-MUC3's modular process model integrates the results of case memor y retrieval over sentences from multiple stories, extracts the date and location of incidents, an d computes cross-reference information for various slots .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "The TTS-MUC3 system incorporates a number of different modules shown below : ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "SYSTEM COMPONENT S", "sec_num": null }, { "text": "The text database provides the capability to retrieve a fragment of text from a large collection tha t may be spread over multiple disk files . In 'ITS-MUC3, the text database was used to store : (1) the database of training stories, (2) the database of testing stories, (3) the database of training template s for user browsing, and (4) the database of parsed templates for use by the associative cas e memory. Retrievals from the text database may return : (1) a raw text string (used for templates) ,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Text Database", "sec_num": null }, { "text": "(2) a recursive token structure with individual words at the leaf nodes (used for stories), or (3) a n s-expression (used for parsed templates) .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Text Database", "sec_num": null }, { "text": "The phrasal parser is a fast, shallow, conceptual parser . The parser accepts a token structure, a lexical hierarchy, and a phrase pattern set. The parser returns an ordered list of text features . A text feature includes: (1) a member of the concept hierarchy, (2) the string covered by the phrase, and (3) a recursive token structure spanning the tokens covered by the phrase .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "Lexicon entries are created by adding word stems to a concept hierarchy as follows ,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "(ks :isa h-lex \"PRIEST\" :religious-individual-w) (ks :isa h-lex \"MISSIONARY\" :religious-individual-w ) (ks :isa h-lex \"CONFERENCE\" :conference-w ) (ks :isa h-lex \"SUMMIT\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": ":conference-w ) (ks :isa h-lex \"RECEPTION\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": ":conference-w )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "Phrasal patterns may reference either elements of the concept hierarchy, or specific words :", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "(ph :defpattern (net ? h-con) ( :determiner :small-numbe r unidentified-w :human-group-w ) civilian ) (ph :defpattern (net ? h-con) ( :civilian-w \"FROM\" :number-w spanish-name-w \"AREA\" ) civilian ) (ph :defpattern (net ? h-con) ( :public-w :communication-device-w building-w ) communications )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "The features are extracted using a depth first search of the patterns, with a preference for pattern s that have specific words over those the have only concept names and a preference for longe r patterns.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Phrasal Parse r", "sec_num": null }, { "text": "The case memory takes an ordered list of text features and returns the K-nearest neighbors . For TTS-MUC3, K was 12 and the metric was the Euclidean distance in a binary vector space . The case memory also accepts a set of slots to fill (set fill and string fill) . For each sentence, the cas e memory returns weighted suggestions for filling each of the requested slots . Case indices are kept in main memory . Parsed templates, used for computing slot fillers, are loaded as needed .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Case Memory", "sec_num": null }, { "text": "The process model has four phases : (1) memory access, (2) topic grouping, (3) slot filling, and (4 ) template generation. There is also an initial training phase which initializes the case memory .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Process Mode l", "sec_num": null }, { "text": "The training phase uses the provided templates to build up the phrase lexicon and the case memory . Phrases are generated from the fillers for the template slots . Cases are generated from th e sentences that provided the fillers . The word lexicon is generated by performing a word frequency analysis on the raw text. For TTS-MUC3, all words that occurred between 10 and 105 times wer e included in the lexicon .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Training Phase", "sec_num": null }, { "text": "For each sentence of a story, the memory access phase queries the case memory to obtai n suggestions for all slots . The resulting structure contains all the weighted suggestions and th e source cases.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Memory Access", "sec_num": null }, { "text": "Topic grouping (analogous to discourse processing) is based on the TYPE OF INCIDENT slot. The weight for each type of incident is computed for every sentence . The weights are then passed through a competitive filter, resulting in binary signals . The competitive filter first normalizes th e topic weights using a Gaussian mask on a sentence by sentence basis, then computes the bes t topic. A topic is a set of contiguous sentences with the same computed value for TYPE OF INCIDENT . Figure 2 shows the inputs and outputs to the topic grouping process . Note that moderately high evidence of kidnapping throughout the story is suppressed in favor of the bombing interpretation , which turns out to be correct . This filter used is topic grouping is designed to pick out signals tha t are high but that \"drop out\" from time to time, as one can see in the smoothing over the arson signal. Slot filling consists of five parts : (1) pure set fills, (2) string fills, (3) cross-referenced slots, (4) date extraction, and (5) location extraction. The first three parts consider only relevant sentences . A relevant sentence shares the same topic with the previous sentence or contains no competin g topic. There are two distinct types of processing for slot filling. Most slots are filled using hypotheses returned by the associative memory, two, date and location, are filled by domai n specific procedures. Three types of slots are filled from the associative memory :", "cite_spans": [], "ref_spans": [ { "start": 486, "end": 494, "text": "Figure 2", "ref_id": "FIGREF1" } ], "eq_spans": [], "section": "Topic grouping and relevance assessmen t", "sec_num": null }, { "text": "1.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Topic grouping and relevance assessmen t", "sec_num": null }, { "text": "Set fills-Pure set fills are computed by averaging the weights over all sentence s for a given topic and picking the highest score .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Topic grouping and relevance assessmen t", "sec_num": null }, { "text": "String fills-String fills are computed in a similar manner to set fills . The difference is that the suggestions returned by the case memory are subject to a threshold on the weights . For the official run of TTS-MUC3, the string fil l threshold was set at 0.1 .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "2.", "sec_num": null }, { "text": "Cross reference generation-Cross reference generation is performed by choosing the most likely tag (as suggested by the case-memory) for the sentence that contain s the string fill.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.", "sec_num": null }, { "text": "For date extraction, all sentences within a topic are scanned for absolute or relative date references . Absolute date references are combined into a range . Absolute dates are preferred over relativ e dates within a given sentence . Relative date references are interpreted with respect to either th e current date specification for a story (if one has been found) or the story date line .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Date processing", "sec_num": null }, { "text": "For location extraction, all sentences within a topic are scanned for known location names . The resulting list of location names is then searched for a maximal, legal, location containment chain .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Location processing", "sec_num": null }, { "text": "For the first sentence in TST-MUC3-0099 ,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "EXAMPLE RUN", "sec_num": null }, { "text": "TTS-MUC3 extracts the following features , ((feature :police-w \"POLICE\" # ) (feature :statement-w \"REPORTED\" # ) (feature :terrorist-act-indiv \"TERRORISTS \" # ) (feature :time-of-day-w \"TONIGHT\" # ) (feature :explosive-w \"BOMBED\" # ) (feature :embassy-w \"EMBASSIES\" # ) (feature :place-name \"PRC\" # ) (feature :place-name \"SOVIET UNION\" #) )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "Based on semantic features such as : POLICE-W and : PLACE-NAME, the following template (along with approximately 11 others) is retrieved from memory . Stored along with the story are the features extracted from DEV-MUC3-0174 that were used t o index the template.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "((feature (feature :police-w \"POLICE\" ) (feature :statement-w \"BELIEVE\" ) (feature :explosive-w \"BOMB\" ) (feature :depart-w \"GOING\" ) (feature :place-name \"U .S .\" ) (feature :embassy-w \"EMBASSY\" ) (feature :commercial-target-w \"WAREHOUSE\" ) (feature :human-individual-w \"MEMBERS\" ) (feature :terrorist-act-org \"THE HONDURAN LEFT\" ) (feature :civilian-w \"PEOPLE\" ) (feature :month-name-w \"APRIL\" ) (feature :favoring-w \"FOR\" ) (feature :place-name \"UNITED STATES\") )", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "Comparing the strings in the retrieved template with the strings for the indexing features, TTS-MUC3 looks for a feature in the new sentence that matches the features (FEATURE : EMBASSY-W \"EMBASSY\") . Using the semantic feature, :EMBASSY-W, TTS-MUC3 proposes (FEATURE : EMBASSY-W \"EMBASSIES\" #) , as a hypothesis for the physica l target in the new story.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "Processing proceeds in a like manner for the rest of the story to produce the following template , TST1-MUC3-009 9 1 25 OCT 198 9 BOMBING TERRORIST AC T * \"SHINING PATH \" CLAIMED OR ADMITTED : \"SHINING PATH \" \"PRC EMBASSY \" \"CAR \" \"VEHICLES \" \"USSR EMBASSY \" PLURAL DIPLOMAT OFFICE OR RESIDENCE : \"USSR EMBASSY \" OTHER : \"VEHICLES \" OTHER : \"CAR \" DIPLOMAT OFFICE OR RESIDENCE :", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "\"PRC EMBASSY \" * * * * * PERU : LIMA (DEPARTMENT ) SOME DAMAGE : \"USSR EMBASSY \" SOME DAMAGE : \"VEHICLES \" SOME DAMAGE : \"CAR \" SOME DAMAGE : \"PRC EMBASSY \" * TTS-MUC3 produces a reasonably good fill for this template . Three features are worth noting .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "First, the string fills \"PRC EMBASSY\" and \"USSR EMBASSY\" are extracted from sentence s after the introductory sentence , \"A CAR-BOMB EXPLODED IN FRONT OF THE PRC EMBASSY, WHICH IS I N THE LIMA RESIDENTIAL DISTRICT OF SAN ISIDRO . MEANWHILE, TWO BOMBS WERE THROWN AT A USSR EMBASSY VEHICLE THAT WAS PARKED IN FRONT OF THE EMBASSY LOCATED IN ORRANTIA DISTRICT, NEA R SAN ISIDRO . \"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "The second feature worth noting is that \"CAR\" is picked up as a target, even though it is actually a part of the instrument \"CAR-BOMB\" . The reason for this mistake is a deficiency in the phrase s that pick out semantic features .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "The third feature is that TTS-MUC3 produced only one template where there should have been tw o bombings . This merging of templates with the same incident type is an inevitable result of th e topic grouping used in TTS-MUC3 .", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "\"[TEXT] POLICE HAVE REPORTED THAT TERRORISTS TONIGHT BOMBE D THE EMBASSIES OF THE PRC AND THE SOVIET UNION . \"", "sec_num": null }, { "text": "To test the sensitivity to different training sets, we loaded the associative memory with differen t templates from the development corpus . To show the difference in performance, Table 1 show s the overall recall and performance for the MATCH/MISSING row of the scoring, with variou s portions of the training data loaded . Whenever a training set is loaded, the number of case with a given incident type is limited to prevent sampling bias. For Table 1 the maximum cases per topic is 10. Note that these training sets are much smaller than the full 1200 stories in the DEV corpus , and therefore the recall performance is substantially lower than the 31% achieved with the ful l training set on TST2. Table 2 : Recall and precision for various training sets with 4 cases per incident type Table 2 presents results similar to Table 1 but with a maximum of four (4) cases per topic . Intuitively, one would imagine that recall at least would fall drastically . Table 2 confirms that intuition as, for all but one training set, the recall drops when fewer cases per incident type ar e loaded . Both Tables 1 and 2 are the result of running the first 10 stories in the TST1 corpus through TTS-MUC3. The first ten stories contain two ARSON templates, and even after limiting the number of cases per topic to 10, ARSON still has fewer than half as many cases as the mor e common types : ATTACK, MUDER, BOMBING, and KIDNAPPING . However, when the numbe r of cases per topic is limited to four (4), ARSON is perfectly balanced with the others . This under representation of ARSON in the training data may account for the anomaly between Tables 1 and 2 for stories 626-700.", "cite_spans": [], "ref_spans": [ { "start": 180, "end": 187, "text": "Table 1", "ref_id": "TABREF2" }, { "start": 447, "end": 454, "text": "Table 1", "ref_id": "TABREF2" }, { "start": 703, "end": 710, "text": "Table 2", "ref_id": null }, { "start": 791, "end": 798, "text": "Table 2", "ref_id": null }, { "start": 827, "end": 834, "text": "Table 1", "ref_id": "TABREF2" }, { "start": 961, "end": 968, "text": "Table 2", "ref_id": null }, { "start": 1631, "end": 1645, "text": "Tables 1 and 2", "ref_id": "TABREF2" } ], "eq_spans": [], "section": "SENSITIVITY TO TRAINING SET", "sec_num": null } ], "back_matter": [ { "text": "To understand the performance of TTS-MlUC3, one should look at the the inter-dependenc e between the various processing modules. ] Figure 3 shows these dependencies . Each modul e points to the modules it depends on . Our contention is that improving a module will enable improvement of the behavior of its dependents .For example, the case memory alone has recall and precisio n rates above 50% . Subsequen t processing results in information loss that accounts for our final rates of 31% and 36% , respectively .", "cite_spans": [], "ref_spans": [ { "start": 131, "end": 139, "text": "Figure 3", "ref_id": null } ], "eq_spans": [], "section": "SUMMAR Y", "sec_num": null }, { "text": "We believe that this ability to analyze, from a system wide perspective, where the errors occur i s unique to TTS . From Figure 3 , we can see that even a perfect case memory would not completely solve all performance problems, as every other component depends on topic grouping . Therefore we conclude that topic grouping is the system component where the most work is needed . We might also deduce that in topic group, we will find the largest leverage for adding knowledge to th e processing . This conclusion concurs with conventional wisdom in natural language, that understanding text across sentence boundaries requires more knowledge that understanding within a sentence .", "cite_spans": [], "ref_spans": [ { "start": 121, "end": 129, "text": "Figure 3", "ref_id": null } ], "eq_spans": [], "section": "Figure 3 : Module Dependency Graph", "sec_num": null } ], "bib_entries": {}, "ref_entries": { "FIGREF0": { "uris": null, "num": null, "type_str": "figure", "text": "Figure 1 : Hughes 'ITS System Block Diagra m" }, "FIGREF1": { "uris": null, "num": null, "type_str": "figure", "text": "Input and output to topic grouping for TST-MUC3-0099 Slot filling" }, "FIGREF2": { "uris": null, "num": null, "type_str": "figure", "text": "TARGET : ID(S ) 12. HUMAN TARGET : TOTAL NUM 13. HUMAN TARGET : TYPE(S ) 14. TARGET : FOREIGN NATION(S) UNITED STATES : \"U .S . EMBASSY WAREHOUSE \" 15. INSTRUMENT : TYPE(s) * 16. LOCATION OF INCIDENT HONDURAS : TEGUCIGALPA (CITY ) 17. EFFECT ON PHYSICAL TARGET(S) NO DAMAGE : \"U .S . EMBASSY WAREHOUSE \" 18. EFFECT ON HUMAN TARGET(S) -" }, "TABREF2": { "content": "
Training StoriesRecallPrecision
1-1001643
101-1751440
476-5501745
551-625836
626-7001945
", "num": null, "text": "Recall and precision for various training sets with 10 cases per incident type", "type_str": "table", "html": null } } } }