File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/05/w05-1613_metho.xml
Size: 26,056 bytes
Last Modified: 2025-10-06 14:09:58
<?xml version="1.0" standalone="yes"?> <Paper uid="W05-1613"> <Title>Natural Language Directed Inference in the Presentation of Ontologies</Title> <Section position="3" start_page="0" end_page="0" type="metho"> <SectionTitle> 2 Realisation from Ontology Axioms </SectionTitle> <Paragraph position="0"> Our current research addresses the problem of presenting parts of OWL DL [McGuinness and van Harmelen, 2004] ontologies in natural language. This will extend existing approaches to generating from simpler DLs (e.g. [Wagner et al., 1999]) by taking into account the fact that in a language like OWL DL a concept is described more by a set of constraints than by a frame-like definition. For instance, the bottom of Figure 1 shows a set of axioms relevant to the concept TemporalRegion in an example ontology. Because there may be a number of axioms providing different facts about a concept, the information cannot in general be presented in a single sentence but requires an extended text with multiple sentences, the overall structure having to be planned so as to be coherent as a discourse. Our work is also different from other work which generates text about individuals described using ontologies [Wilcock, 2003; Bontcheva and Wilks, 2004], in that it presents the ontology class axioms themselves.</Paragraph> <Paragraph position="1"> In this section, we give an example of how complicates the reasoning about appropriate content, by showing that although the that we are developing is relatively simple, it nevertheless complicates decisions about the complexity of what can be presented in a sentence.</Paragraph> <Paragraph position="2"> Given an axiom to be expressed as a sentence (we discuss in section 3 how such axioms are selected), our realisation approach uses rules with simple recursive structural patterns and assembles text with grammatically-annotated templates1.</Paragraph> <Paragraph position="3"> The idea is that we will collect rules for special case expressions that can be realised relatively elegantly in language as well as having generic rules that ensure that every possible structure can be handled. Optimal English will arise from detecting the part of speech of any class and role names which are English words (as well as cases such as multiple word names and roles such as &quot;hasX&quot;, &quot;Xof&quot; where X is a noun), and we have been able to obtain this information with reasonable quality automatically using WordNet2. Unless such conventions are used in the ontology definition or the reader is familiar with some of the ontology terms, it will not be possible to convey any useful information to them without extra 1Note that our initial approach is to see how much can be achieved with no restrictions on the ontology (as long as it is expressed in legal OWL DL) and only generic linguistic resources (such as WordNet [Miller, 1995]). This is partly because there is a need to present parts of current ontologies, which often come with no consistent commenting or linguistic annotations, and partly so that we can then make informed recommendations about what kinds of extra annotations would be valuable in the ontologies of the future. Also, note that the term &quot;realisation&quot; will be taken here to include elements of &quot;microplanning&quot; which, for instance, introduces appropriate pronominalisation.</Paragraph> <Paragraph position="4"> 2We cannot guarantee that the ontology writer will use such mnemonic names (if not, then generation will have to use the less optimised templates), but we should exploit these cases when they arise (and our investigations have shown that they are extremely common in human-written ontologies).</Paragraph> <Paragraph position="5"> domain-specific resources.</Paragraph> <Paragraph position="6"> A given axiom may match multiple rules and therefore have multiple possible realisations. For instance, the axiom:</Paragraph> <Section position="1" start_page="0" end_page="0" type="sub_section"> <SectionTitle> Student Person u9Supervisor:Academic </SectionTitle> <Paragraph position="0"> would be mapped to &quot;A student is a person with at least one academic supervisor&quot;, which exploits knowledge of the lexical categories of the names used, but another possibility would be something like &quot;Something in class student is something in class person with at least one value for the role supervisor which is something in class academic&quot; (this might have been the only possibility if the class names had been arbitrary identifiers such as &quot;Class1&quot; and &quot;Class2&quot;.).</Paragraph> <Paragraph position="1"> Where a logical formula has multiple realisations, a measure of linguistic complexity of the results can be used to select a preferred one. Currently we measure linguistic complexity as the number of words in the English string generated. Better measures will take into account the shape of the parse tree. Notice that linguistic complexity does not directly mirror the complexity of the formula, but depends on and whatever linguistic resources underlie it. Although more complex formulae tend to yield more complex linguistic output, linguistic complexity is also affected by: The extent to which special-case shorter rules match some of its subexpressions The extent to which class and role names can be interpreted as English words of relevant classes Whether a recursive linguistic structure uses left, right or centre embedding [Miller and Isard, 1964] The linguistic complexity of a formula is obtained by taking the linguistic complexity of the realisation that is least complex. Again, although there is a correlation with the complexity of the formula, the relevant complexity for deciding, for instance, whether a formula can be presented in a single sentence, is a linguistic one which needs to take into account.</Paragraph> <Paragraph position="2"> What is a Temporal Region? [One kind of Temporal Region is a Time Interval.] [A Perdurant can happen at a Time Interval.] [nothing is both a Temporal Region and an Abstract Region.]but[An Abstract Region is also a kind of Region] [A Temporal Region is a kind of Region.]</Paragraph> </Section> </Section> <Section position="4" start_page="0" end_page="0" type="metho"> <SectionTitle> 3 Selecting Material </SectionTitle> <Paragraph position="0"> The designer of an ontology has chosen one of many possible logically equivalent ways to axiomatise their domain, and this is important information. Therefore our initial approach worked from the axioms themselves without manipulating them in any way.</Paragraph> <Paragraph position="1"> We basically followed the same procedure for content determination as in the ILEX system [O'Donnell et al., 2001]. Thus the axioms can be seen as forming a graph, where each axiom is connected to the concepts it mentions (and where there may also be other links for relations between axioms) - see Figure 1. In this graph, routes between axioms correspond to different possible transitions in a coherent text a text proceeds from one sentence to another by exploiting shared entities or by virtue of a rhetorical relation between the sentences3.</Paragraph> <Paragraph position="2"> A possible hand-generated text from the above axioms, showing the coherence relations which hold by virtue of shared entities or a rhetorical relation (the latter shown in dashes) is shown in Figure 2.</Paragraph> <Paragraph position="3"> Assuming for the moment that a user has asked the question What is X?, where X is some class used in the ontology, selecting the axioms to express in the answer involves a best-first search for axioms, starting at the entity X. Each axiom is evaluated according to: how close it is (in terms of edges of the graph) to the concept X, and how intrinsically interesting, important and understandable it is.</Paragraph> <Paragraph position="4"> how few times is has already been presented Following ILEX, these three measures are multiplied together and, for a text of length n, the n facts with the highest measures are selected for inclusion. The first component of the measure ensures that the retrieved axioms are relevant to the question to be answered. In terms of this, the best axioms to use are ones directly involving the class X. On the other 3The ILEX model makes use of the idea that there may be rhetorical relations, such as concession or background, between facts which could potentially be expressed in a text. It is however not immediately clear how they arise in our context. It seems that it may be plausible to say, for instance, &quot;Although students are people and lecturers are people, lecturers and students are disjoint&quot;, but the general principles for this need to be worked out.</Paragraph> <Paragraph position="5"> hand, axioms that are only indirectly involved with X can be selected if they score well according to the second component (or if there are not enough closer axioms). The fact that there is a path between X and each chosen axiom ensures that there is a way of linking the two in a coherent text, by progressively moving the focus entity of the text to new entities in the axioms already expressed or through expressing rhetorical relations.</Paragraph> <Paragraph position="6"> The second component of the evaluation score for axioms can be used to make the system sensitive to the user, for instance by preferring axioms that involve concepts known to the user or axioms that have not previously been told to them. We have not yet exploited this feature. The third component penalises axioms that have already been presented.</Paragraph> </Section> <Section position="5" start_page="0" end_page="0" type="metho"> <SectionTitle> 4 Natural Language Directed Inference </SectionTitle> <Paragraph position="0"> The content determination approach just described, which selects from among the provided axioms, suffers from a number of deficiencies: Over-complex sentences: The axioms may not package the available information appropriately for natural language sentences. On the one hand, an axiom may be too complex to express in a single sentence (as determined by applying and measuring the linguistic complexity). In this case, it might be appropriate to present a &quot;weaker&quot; (axiom). For instance, instead of expressing X Y t Z t : : : one might express Y v X (if it mentions the entities needed for coherence with the rest of the text).</Paragraph> <Paragraph position="1"> Repetitive sentences: On the other hand, the axioms may give rise to sentences that are short and repetitive. Thus, rather than using three sentences to express:</Paragraph> <Section position="1" start_page="0" end_page="0" type="sub_section"> <SectionTitle> Student v Person Student v UnEmployed Student v 9Supervisor:Academic </SectionTitle> <Paragraph position="0"> one could combine them all into a formula realised as &quot;a student is an unemployed person with at least one academic supervisor&quot;. In NLG, the process of building such complex sentences is known as &quot;aggregation&quot; [Shaw, 1995]. This kind of aggregation could be implemented by combining the axioms together before realisation is performed, but success can only be measured by looking at the linguistic complexity of the result.</Paragraph> <Paragraph position="1"> Inappropriate focus: An axiom may be expressed in a way that, when realised, places inappropriate emphasis on entities. For instance, an axiom X v Y could be realised by &quot;An X is a kind of Y&quot;, whereas the equivalent Y w X could be realised by &quot;Y's include X's&quot;. The latter would be much better than the former at a point in a text that is discussing the properties of Y. The above example of &quot;weakening&quot; also has the effect of changing the likely subject of the sentence produced. Sometimes the text will be better if one can switch around the material in an axiom to emphasise different material.</Paragraph> <Paragraph position="2"> Misleading partial information: It may be better to present some of the consequences of an axiom, given the rest of the theory, rather than the axiom itself. For instance, instead of presenting Student v 9supervisor:Academic in an ontology which also has the axiom functional(supervisor), it would be more informative to present the consequence Student v = 1 supervisor:Academic Indeed, with number restrictions a reader can draw false implicatures (in the sense of [Grice, 1975]) if only partial information is presented. In this case, a scalar implicature [Levinson, 1983] is involved. A reader, on being told that &quot;a student has at least one academic supervisor&quot;, will naturally assume that they could have more than one, or that they could have other supervisors belonging to other classes. Similarly, on being told &quot;a supervisor of a student is always an academic&quot;, one will assume that there can be more than one supervisor (otherwise the text would have said &quot;the supervisor : : :&quot;). Some of the principles at work here may be similar to those encountered in cooperative question answering [Gaasterland et al., 1992].</Paragraph> <Paragraph position="3"> The only way to overcome these limitations is to enable content determination to select material in more ways than just choosing an axiom. It must always choose to express something that is true, given the logical theory, and content determination will therefore be a form of inference. In general, in fact, we could consider using any logical consequence of the axioms. However, not all logical consequences are equally good. The formulae that are presented should: 1. Soundness: follow from the original logical theory (set of axioms) 2. Relevance: contribute information relevant to the goal of the text. For instance, if the goal is to answer the question &quot;what is concept X?&quot; then the formulae should be about X or other concepts which shed light on X.</Paragraph> <Paragraph position="4"> 3. Conservatism: be not very different from the original axioms (and so capture some of the intent behind those axioms) 4. Complexity: have appropriate linguistic complexity (section 2) 5. Coherence: satisfy linguistic coherence constraints (i.e. be linked to other selected material by the kinds of relations discussed in section 3).</Paragraph> <Paragraph position="5"> 6. Novelty: not have already been expressed (and not be tautologies). There is no point in weakening axioms to the point that nothing new is expressed, or in presenting the same material many times.</Paragraph> <Paragraph position="6"> 7. Fullness: be complete, to the extent that they don't support false implicatures 8. User-orientation: be in accord with user model prefer null ences (as in section 3) We call the kind of inference required to find such logical consequences natural language directed inference (NLDI). It is a kind of forwards inference with very specific goals, which arise from its use for natural language generation.</Paragraph> <Paragraph position="7"> Although we have motivated NLDI through our own particular content determination problem, this may be a useful way to view content determination in general, as long as the starting point can be viewed as some kind of logical theory, , there is an available realisation relation and an evaluation function eval for linguistic outputs, which takes into account the above desiderata. In this case, content determination can be viewed as the problem of determining argmax( such that j= ) maxfeval(t)jt 2 ( )g The process of enumerating promising consequences of for this optimisation is certainly a form of logical inference. But its goal is unlike standard goals of automated reasoning and is shaped by the idiosyncracies of the requirements for natural language output. There is an interesting parallel here with the work of [Sripada et al., 2003]. Sripada et al found that, for generating natural language summaries of time series data, standard data analysis algorithms such as segmentation had to be modified. They characterised the extra requirements that forced these modifications in terms of the Gricean maxims of cooperative communication [Grice, 1975]. Our 8 desiderata above could also be thought of as cases of the Gricean maxims.</Paragraph> </Section> </Section> <Section position="6" start_page="0" end_page="0" type="metho"> <SectionTitle> 5 Techniques for NLDI </SectionTitle> <Paragraph position="0"> Unfortunately, standard refutation-based approaches to inference rely on having a precisely specified inference goal, whose negation is incompatible with the axioms. For DLs, the standard tableaux methods [Horrocks, 1998] have similar properties. NLDI does not have an inference goal that can be expressed in structural terms, so even approaches to &quot;matching&quot; cannot straightforwardly be used to derive linguistically appropriate results. NLDI is more akin to other &quot;non-standard&quot; types of inference, perhaps to approximation [Brandt et al., 2002], though again the target logical language is without a simple formal characterisation. Perhaps the closest approach we are aware of is meta-level control of inference, where factors outside of the logic (e.g. other kinds of descriptions of the shapes of logical formulae) are used to guide inference [Bundy and Welham, 1981].</Paragraph> <Paragraph position="1"> One advantage of NLDI is that it does not have to be a complete inference procedure, though in general the more logical consequences of the original axioms it can find, the more possible texts will be considered and the higher the quality of the one chosen.</Paragraph> <Paragraph position="2"> The approach to NLDI we are currently working on is inspired by the idea of &quot;overgeneration&quot; approaches to NLG, as used, for instance, by those using statistical models [Langkilde and Knight, 1998] and instance-based search [Varges and Mellish, 2001]. In this approach, instead of attempting to intelligently order the relevant choices to come up with an optimal text, an NLG system consciously enumerates a large number of possible texts (in a cheap way) and then chooses between them using a linguistically-aware evaluation function of some kind (the eval of NLDI). Our approach differs from these others, however, in that, whereas the other systems implement overgeneration of surface forms, we consider overgeneration of possible content.</Paragraph> <Paragraph position="3"> Figure 3 shows the architecture of our system under development. The simple inference system implements a beam search among possible sets of content for generating texts, where each state in the search space is a sequence of formulae. In logical terms, each sequence represents a conjunction that follows from the input axioms. The resulting text for any such sequence (i.e. the result of applying ) will be the result of realising the elements of the sequence, in order, as the sentences of the text.</Paragraph> <Paragraph position="4"> At each point in the search, the current state can give rise to new states in two possible ways: 1. One of the original axioms is added to the end of the sequence.</Paragraph> <Paragraph position="5"> 2. The final formula of the sequence is replaced by a for- null mula inferred from it (given the whole axiom set) by one inference step The inference steps represent simple ways of modifying a formula to something close to it which follows from the complete set of axioms and which may yield a more appropriate realisation. We have currently implemented a small number of relevant steps, including steps for aggregation, disaggregation and elimination of disjunctions.</Paragraph> <Paragraph position="6"> Whenever a new state in the search space is generated, it is sent to the realisation component (which implements ) and from there through an evaluation function (which implements eval). The evaluation function takes into account the average deviation of the sentence lengths (in words) from an &quot;ideal&quot; sentence length and some other heuristics (see below). This is used as feedback to drive the search of the inference component in a best-first manner. The search terminates when the best scoring state is one element longer that the desired number of sentences for the text, at which point its sequence of formulae, apart from the last one, is returned. The exploration to a length longer than the desired one ensures that other states shorter than or equal to the desired length have a chance to be explored.</Paragraph> <Paragraph position="7"> Our approach makes initial attempts to address the desider- null ata of NLDI by constraining the search in the following ways: 1. Soundness: All new formulae are derived by sound rules of inference from the existing axioms and so are true.</Paragraph> <Paragraph position="8"> 2. Relevance: Only axioms which might affect the interpretation of the class asked about are ever considered (the rest are discarded at the start of the process). For the purposes of this, we use the conservative relevant-only translation of [Tsarkov et al., 2004] to discard axioms that cannot be relevant to the question.</Paragraph> <Paragraph position="9"> 3. Conservatism: Inferred formulae are based on individual axioms, and shorter inferences are enumerated before longer ones.</Paragraph> <Paragraph position="10"> 4. Complexity: The complexity of the best realisations is used to order the search candidates. Candidates which are inappropriate for realisation do not match the realisation rules and so are not considered.</Paragraph> <Paragraph position="11"> 5. Coherence: When a new axiom is added to a sequence, it is constrained in its realisation to have a subject which is a class mentioned in the previous element of the sequence. The subject of the first element of the sequence must be the class which is the subject of the original question. Also the evaluation function has a preference for the first sentence with a given class as subject to be an &quot;is a&quot; type sentence.</Paragraph> <Paragraph position="12"> 6. Novelty: In order to prevent information being presented more than once, only one logical consequence of any given axiom is ever included in a sequence. This is implemented via a simple way of tracking the axioms that have contributed to each formula. This makes the assumption that the original axioms are logically independent. null 7. Fullness: Formulae are closed with respect to cardinality information before being added to the lists.</Paragraph> <Paragraph position="13"> 8. User-orientation: We don't currently take this into account, but intend to reward formulae that contain class and role names already familiar to the user (e.g. used in answers to previous questions, or appearing earlier in the answer to the current question).</Paragraph> <Paragraph position="14"> All of these are relatively crude measures, which nevertheless give some appropriate direction to the process.</Paragraph> <Paragraph position="15"> This system has been implemented and tested informally on examples from three different ontologies. For example, in creating a 3-sentence text to answer &quot;What is an Electrode?&quot; using a fuel cell ontology with 133 axioms, the relevance filter first of all reduces the set of axioms to 31, which include: (1) Electrode v Actuality (2) Electrode v9contains:Catalyst (3) Electrode v (9contains:Support u 1 contains:>) (4) domain(contains; FuelCell t MEA t Electrode t Catalyst) as well as other axioms such as Catalyst v 9contains:ActiveMetal. If these 4 axioms were selected unchanged and realised in this order (which by chance happens to be quite a good order), then the following text would result: An Electrode is a kind of Actuality. An Electrode always contains something which is a Catalyst. An Electrode always contains something which is a Support and always contains at most 1 thing. Only something which is a FuelCell, a MEA, an Electrode or a Catalyst contains something.</Paragraph> <Paragraph position="16"> Instead of this, our simple implementation of NLDI proceeds as follows. The initial states are those axioms which when realised will have Electrode in subject position, with a preference for those that will be realised as &quot;an Electrode is a ...&quot;. Thus the state consisting of the one element sequence: Electrode v Actuality will be a favourite. This state can be developed in several ways. For instance, another axiom could be aggregated with this one (to give a sentence of the form &quot;an Electrode is an actuality which ...&quot;). Another possibility is for this axiom to be accepted in this form and for another axiom to be added to the end of the sequence. This second possibility generates the following state, among others: Electrode v Actuality Electrode v = 1 contains:Catalyst (notice how more precise cardinality information has been attached to axiom (2)). This state can be further developed by adding a further axiom, or by applying an inference rule to the last added formula. In this case, aggregation with axiom This state is further developed by adding new axioms to the end, and so on. The final sequence of formulae selected is: Electrode v Actuality Electrode v = 1 contains:(CatalystuSupport) domain(contains; FuelCell t MEA t Electrode t Catalyst) This is realised by the following short text: An Electrode is a kind of Actuality. An Electrode contains exactly one thing, which must be a Catalyst and a Support. Only something which is a FuelCell, a MEA, an Electrode or a Catalyst contains something.</Paragraph> <Paragraph position="17"> (This realisation relies on part-of-speech information which can be obtained automatically from WordNet, apart from the term &quot;MEA&quot;).</Paragraph> </Section> class="xml-element"></Paper>