{ "paper_id": "P82-1016", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T09:16:12.998769Z" }, "title": "An Improved Heuristic for Ellipsis Processing*", "authors": [ { "first": "Ralph", "middle": [ "M" ], "last": "Welschedel", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Delaware Newark", "location": { "postCode": "19711", "settlement": "Delaware" } }, "email": "" }, { "first": "Norman", "middle": [ "K" ], "last": "Sondheimer", "suffix": "", "affiliation": { "laboratory": "Software Research Sperry Univac MS", "institution": "", "location": { "addrLine": "2G3 Blue Bell", "postCode": "19424", "region": "Pennsylvania" } }, "email": "" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "", "pdf_parse": { "paper_id": "P82-1016", "_pdf_hash": "", "abstract": [], "body_text": [ { "text": "Robust response to ellipsis (fragmentary sentences) is essential to acceptable natural language interfaces. For instance, an experiment with the REL English query system showed 10% elliptical input (Thompson, 1980) .", "cite_spans": [ { "start": 198, "end": 214, "text": "(Thompson, 1980)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": null }, { "text": "Quirk , et al. (1972) One general strategy has been to substitute fragments into the analysis of the previous input, e.g., substituting parse trees of the elliptical input into the parse trees of the previous input in LIFER (Hendrix, et al., 1978) . This only applies to inputs of the same type, e.g., repeated questions.", "cite_spans": [ { "start": 6, "end": 21, "text": ", et al. (1972)", "ref_id": null }, { "start": 224, "end": 247, "text": "(Hendrix, et al., 1978)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "In", "sec_num": null }, { "text": "Allen (1979) deals with some examples of expansion ellipsis, by fitting a parsed elliptical input into a model of the speaker's plan. This is similar to other methods that interpret fragments by placing them into prepared fields in frames or case slots (Schank et al., 1980; Hayes and Mouradian, 1980; Waltz, 1978) . This approach seems most applicable to limiteddomain systems.", "cite_spans": [ { "start": 253, "end": 274, "text": "(Schank et al., 1980;", "ref_id": null }, { "start": 275, "end": 301, "text": "Hayes and Mouradian, 1980;", "ref_id": null }, { "start": 302, "end": 314, "text": "Waltz, 1978)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "In", "sec_num": null }, { "text": "The Heuristic There are three aspects to our solutien: a mechanism for repetition and replacement ellipsis, an extension for inputs of different types, such as fragmentary answers to questions, and an extension for expansion ellipsis.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.", "sec_num": null }, { "text": "As noted above, repetition and replacement ellipsis can be viewed as substitution in the previous form.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Repetition and Replacement", "sec_num": "3.1" }, { "text": "We have implemented this notion in an augmented transition network (ATN) grammar interpreter with the assumption that the \"previous form\" is the complete ATN path that parsed the previous input and that the lexical items consumed along that path are associated with the arcs that consumed them.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Repetition and Replacement", "sec_num": "3.1" }, { "text": "In ellipsis mode, the ATN interpreter executes the path using the elliptical input in the following way:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Repetition and Replacement", "sec_num": "3.1" }, { "text": "I.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Repetition and Replacement", "sec_num": "3.1" }, { "text": "Words from the elliptical input, i.e., the curren~ input, may be consumed along the path at any point.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Repetition and Replacement", "sec_num": "3.1" }, { "text": "Any arc requiring a word not found in the current input may be traversed using the lexical item associated with the arc from the previous input. 3.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "2.", "sec_num": null }, { "text": "However, once the path consumes the first word from the elliptical input, all words from the elliptical input must be consumed before an arc can use a word from the previous input.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "2.", "sec_num": null }, { "text": "Traversing a PUSH arc may be accom ~ plished either by following the subpath of the previous input or by finding any constituent ef the required type in the current input. The entire ATN can be used in these cases.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "Suppose that the path for \"Were you angry?\" is given by ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "(CAT COPULA ... (TO Sx)) \"w--~'r~e\" Sx (PUSH NP ... (TO Sy)) [NP (CAT PRO ... (TO NPa)) \"you\" NPa (POP ...) ] Sy", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "(CAT ADJ ... (TO Sz)) \"angry\" Sz (POP ...) Table I An ATN Path for \"Were you Angry?\" An elliptical input of \"Was he?\" following \"Were you angry?\" could be understeed by traversing all of the arcs as in Table I .", "cite_spans": [], "ref_spans": [ { "start": 43, "end": 50, "text": "Table I", "ref_id": "TABREF2" }, { "start": 202, "end": 209, "text": "Table I", "ref_id": "TABREF2" } ], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "Following point I above, \"was\" and \"he\" would be substituted for \"were\" and \"you\". Following point 3, in traversing the arc (CAT ADJ ... (TO Sz)) the lexical item \"angry\" from the previous input would be used. Item 4 is illustrated by an elliptical input of \"Was the old man?\"; this is understood by traversing the arcs at the S level of Our technique is to restrict the mapping such that any expected parse path is generated by applying only one transformation and applying it only once. A special feature of our transformational system is the automatic allowance for dialogue diexis.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "An expected parse path for the answer to \"Were you angry?\" is given in Table 2 . Note in Table 2 , \"you\" has become \"I\" and \"were\" has become \"was\" \"was \" Sy \"angry\" Sz Table 2 Declarative for the expected answer for \"Were you angry?\".", "cite_spans": [], "ref_spans": [ { "start": 71, "end": 78, "text": "Table 2", "ref_id": null }, { "start": 89, "end": 96, "text": "Table 2", "ref_id": null }, { "start": 169, "end": 176, "text": "Table 2", "ref_id": null } ], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "Using this path, the ellipsis interpreter de'scribed in Section 3.1 would understand the ellipses in \"a)\" and \"b)\" below, in the same way as \"a')\" and \"b'i\" a) I was. Probably (I was angry).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.", "sec_num": null }, { "text": "For a time (I was angry).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "(PUSH PF ... (To s))", "sec_num": null }, { "text": "(* this includes a teat that the NP is one of time or place) \u2022 .. (TO S)) Yesterday (I was angry).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "(PUS~ ~P", "sec_num": null }, { "text": "... (TO Sy)) (I was) very (angry).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "(PUSH INTENSIFIER-ADVERB", "sec_num": null }, { "text": "Since this is an extension to the ellipsis interpreter, combinations of repetition, replacement, and expansion can all be handled by the one mechanism. For instance, in response to \"Were you angry?\", \"Yesterday you were (angry)\" would be treated using the expansion and replacement mechanisms. The o~her constraint is that the input must be consumed as a contiguous string.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Table 3 Example Expansion Paths", "sec_num": null }, { "text": "This constraint is violated, for instance, in \"I was (angry) yesterday\" as a response to \"Were you angry?\" Nevertheless, the constraint is computationally useful, since allowing arbitrary gaps in consuming the elliptical input produces a very large space of correct interpretations.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "A ludicrous example is the following question and elliptical response:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Has the boss given our mutual friend a raise? A fat raise.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Allowing arbitrary gaps between the substrings of the ellipsis allows an interpretation such as \"A (boss has given our) fat (friend a) raise.\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "While it may be possible to view all contextual ellipsis as combinations of the operations repetition, replacement, and expansion applied to something, our model makes the strong assumption that these operations may be viewed as applying to an ATN path rather straightforwardly related to the previous utterance.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Not all expansions can be viewed that way, as example f in Section I illustrates. Also, answers of \"No\" require special processing; that response in answer to \"Were you angry\" should not be interpreted as \"No, I was angry.\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "One should be able to account for such examples within the heuristic described in this paper, perhaps by allowing the transformation system described in section 3.2 to be completely general rather than strongly restricted to one and only one transformation application.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Rowever, we propose handling such cases by special purpose rules we are developing. These rules for the special cases, plus the mechanism described in section 3 together will be formally equivalent in predictive power to a grammar for elliptical forms.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Though the heuristic is independent of the individual grammar, designating expansion paths and transformations obviously is not.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "The grammar may make this an easy oz\" difficult task. For instance in the grammar we are using, a subnetwork that collects all tense, aspect, and modality elements would simplify some of the transformations and expansion paths.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "~aturally, semantics must play an important part in ellipsis processing. Consider the utterance pair below:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "Did the bess have a martini at lunch? Some wine.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Special Cases and Limitations", "sec_num": null }, { "text": "syntactically this could be interpreted either as \"Some wine (did have a martini at lunch)\", \"(The boss did have) some wine (at lunch)\", or \"(The boss did have a martini at) some wine\". Semantics should prefer the second reading.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Though", "sec_num": null }, { "text": "We are testing our heuristic using the RUS grammar (Bebrow, 1978) which has frequent calls from the grammar requesting that the semantic component decide whether to build a semantic interpretation for the partial parse found or to veto that partial parse. This should aid performance.", "cite_spans": [ { "start": 51, "end": 65, "text": "(Bebrow, 1978)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Though", "sec_num": null }, { "text": "There are three aspects te our solution:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Summary and Conclusion", "sec_num": null }, { "text": "a mechanism for repetition and replacement ellipsis, an extension for inputs of different types, such as fragmentary answers to questions, and an extension for expansion ellipsis.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Summary and Conclusion", "sec_num": null }, { "text": "Our heuristic deals with the three types of expansion ellipsis as follows: Repetition ellipsis is processed by repeating specific parts of a transformed previous path using the same phrases as in the transformed form (\"I was angry\"). Replacement ellipsis is processed by substituting the elliptical input for contiguous constituents on a transformed previous path.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Summary and Conclusion", "sec_num": null }, { "text": "Expansion ellipsis may be processed by taking specially marked paths that detour from a given state in that path.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Summary and Conclusion", "sec_num": null }, { "text": "Combinations of the three types of ellipsis are represented by combinations of the three variations in a transformed previous path.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "~. Summary and Conclusion", "sec_num": null }, { "text": "are two contributions of the work.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "There", "sec_num": null }, { "text": "First, our method allows for expansion ellipsis. Second, it accounts for combinations of previous sentence form and ellided form, e.g., statement following question, question following statement, question following question.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "There", "sec_num": null }, { "text": "Furthermore, the method works without any constraints on the ATN grammar. The heuristics carry over to formalisms similar to the ATN, such as context-free grammars and augmented phrase structure grammars.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "There", "sec_num": null }, { "text": "Our study of ellipsis is part of a much broader framework we are developing for processing syntactically and/or semantically ill-formed input; see Weischedel and Sondheimer (1981) .", "cite_spans": [ { "start": 147, "end": 179, "text": "Weischedel and Sondheimer (1981)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "There", "sec_num": null } ], "back_matter": [ { "text": "is due to Amir Razi for his programming assistance.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Acknowledgement ~luch credit", "sec_num": null }, { "text": "Allen, James F., \"A Plan-Based Approach to Speech Act Recognition,\" Ph.D. Thesis, Dept. of'Computer Science, University of Toronto, Toronto, Canada, 1979 ", "cite_spans": [ { "start": 123, "end": 153, "text": "Toronto, Toronto, Canada, 1979", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "References", "sec_num": null } ], "bib_entries": {}, "ref_entries": { "FIGREF0": { "uris": null, "text": "... (TO Sa)) (CAT PRO ... (TO NPa)) (PoP ...) (CAT COPULA ... (TO Sy)) (CAT ADJ ... (TO Sz))", "num": null, "type_str": "figure" }, "TABREF2": { "type_str": "table", "text": "", "num": null, "html": null, "content": "
questions following statements, etc.Our
approachto the problem is to write a set
ef transformations whichmaptheparse
path of a question (e.g., Table I) into an
expectedparsepathforadeclarative
response,andtheparse ~path for a de-
clarative intoapathforanexpected
question, etc.
The left-hand side ofatransforma-
tion is a pattern which is matched against
the ATN path ofthepreviousutterance.
Patternelementsinclude literals refer-
ring te arcs, variables which match a sin-
glearc or embedded path, variables which
match zero or mere arcs, and setsefal-
ternatives.It is straightforward to con-
struct a discrimination netcorresponding
toallleft-handsidesfor efficiently
finding what patterns match theATNpath
ofthe previous sentence.The right-hand
side efatransformationisapattern
whichconstructsanexpected path.The
form of the pattern en the right-hand side
isalist of references to states, arcs,
and lexical entries.Such referencescan
bemadethroughitemsmatchedonthe
left-hand side or by explicit construction
ef literal path elements.
, butusingthe
appropriatepathintheNPnetwork to
parsethe old man
3.2Transformations of the Previous Form
Whiletheapproachillustratedin
Section 3.1 is useful in a data base query
environment where ~]lipticalinputtypi-
callyisamodlfication of the previous
query, it does not account forelliptical
statements following questions, elliptical
" } } } }