File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/metho/05/w05-1602_metho.xml
Size: 13,245 bytes
Last Modified: 2025-10-06 14:10:01
<?xml version="1.0" standalone="yes"?> <Paper uid="W05-1602"> <Title>Interactive Authoring of Logical Forms for Multilingual Generation/</Title> <Section position="4" start_page="0" end_page="0" type="metho"> <SectionTitle> 3 Method and Architecture </SectionTitle> <Paragraph position="0"> We now present the system we have implemented, which we have called SAUT (Semantic AUthoring Tool). Our objective is to perform usability studies to evaluate: + How ontological knowledge in the form of concept and relation hierarchies is useful for semantic authoring; + How natural language feedback improves the authoring - and how feedback in two languages modifies the authoring process; + How user interface functionality improves the speed and accuracy of the authoring.</Paragraph> <Paragraph position="1"> The architecture of the system is depicted in Fig. 2. The two key components of the system are the knowledge acquisition system and the editing component. The knowledge acquisition system is used to derive an ontology from sample texts in a specific domain. In the editing component, users enter logical expressions on the basis of the ontology.</Paragraph> <Section position="1" start_page="0" end_page="0" type="sub_section"> <SectionTitle> 3.1 Knowledge Acquisition </SectionTitle> <Paragraph position="0"> For the acquisition of the concepts/relations database, we use two main sources: Verbnet [Kipper et al., 2000] and WordNet [Miller, 1995].</Paragraph> <Paragraph position="1"> We use the information for bootstrapping concept and relation hierarchies. Given sample texts in the target domain, we perform shallow syntactic analysis and extract nouns, verbs and adjectives from the text. Dependency structures for verbs and nouns are also extracted. We currently perform manually anaphora resolution and word sense disambiguation, since automatic methods do not produce accurate enough results. Given the set of nouns and adjectives, we induce the hypernym hierarchy from WordNet, resulting in a tree of concepts - one for each synset appearing in the list of words in the sample texts.1 In addition to the concept hierarchy, we derive relations among the concepts and predicates by using the Verbnet lexical database [Kipper et al., 2000]. Verbnet supplies information on the conceptual level, in the form of selectional restrictions for the thematic roles.</Paragraph> <Paragraph position="2"> These relations allow us to connect the concepts and relations in the derived ontology to nouns, verbs and adjectives. The selectional restrictions in Verbnet refer to the WordNet conceptual hierarchy. In Verbnet, verbs are classified following Levin's classes [Levin, 1993] and thus its representation is easily adjustable with our verb lexicon [Jing et al., 2000], which combined information on argument structure of verbs from Levin, Comlex [Macleod and Grishman, 1995] and WordNet. The rich information on argument structure and selectional restrictions can be automatically adopted to the domain concepts database. Thus, by connecting a concept to a verb, given all the concepts that stand in relation to it in a specific CG (the verb's arguments and circumstantials) - our lexical chooser finds the suitable structure (alternation) to map the CG to a syntactic structure.</Paragraph> <Paragraph position="3"> The outcome of this process is useful in the lexical and syntactic module of the system due to the flexibility it offers to the lexical chooser (a general word can be used instead of a 1Although hypernym relations in WordNet define a forest of trees, we connect all trees with a general node.</Paragraph> <Paragraph position="4"> specific word i.e. vehicles instead of cars, and for the generality of selectional restrictions on verb/adjective arguments. Since there are no Hebrew parallels to WordNet/verbnet, we use a &quot;naive&quot; scheme of translating the English LC to Hebrew, with manual corrections of specific structures when errors are found.</Paragraph> <Paragraph position="5"> Once the knowledge is acquired, we automatically updated a lexical chooser adopted to the domain. The lexical chooser maps the ontological concepts and relations to nouns, verbs and adjectives in the domain.</Paragraph> </Section> <Section position="2" start_page="0" end_page="0" type="sub_section"> <SectionTitle> 3.2 The SAUT Editor </SectionTitle> <Paragraph position="0"> To describe the SAUT editor, we detail the process of authoring a document using the tool. When the authoring tool is initiated, the next windows are presented (see Fig. 3): + Defaults: rules that are enforced by default on the rest of the document. The defaults can be changed while editing. Defaults specify attribute values which are automatically copied to the authored CGs according to their type.</Paragraph> <Paragraph position="1"> + Participants: a list of objects to which the document refers. Each participant is described by an instance (or a generic) CG, and is given an alias. The system provides an automatic identifier for participants, but these can be changed by the user to a meaningful identifier.</Paragraph> <Paragraph position="2"> + Utterances: editing information proposition by proposition. null The system provides suggestions to complete expressions according to the context in the form of popup windows. In these suggestion windows, the user can either scroll or choose with the mouse or by entering the first letters of the desired word, when the right word is marked by the system, the user can continue, and the word will be automatically completed by the system. For example, when creating a new participant, the editor presents a selection window with all concepts in the ontology that can be instantiated. If the user chooses the concept type &quot;Dog&quot; the system creates a new object of type dog, with the given identifier. The user can further enrich this object with different properties. This is performed using the &quot;.&quot; notation to modify a concept with an attribute. While the user enters the instance specification and its initial properties, a feedback text and a conceptual graph in linear form are generated simultaneously. When the user moves to the next line, the new object is updated on the global context view. Each object is placed in a folder corresponding to its concept type, and will include its instance name and its description in CG linear form.</Paragraph> <Paragraph position="3"> In the Utterances panel, the author enters propositions involving the objects he declared in the participants section. To create an utterance, the user first specifies the object which is the topic of the utterance. The user can choose one of the participants declared earlier from an identifiers list, or by choosing a concept type from a list. Choosing a concept type will result in creating a new instance of this concept type. Every instance created in the system will be viewed in the context viewer. After choosing an initial object, the user can add expressions in order to add information concerning this object. After entering the initial object in an utterance, the user can press the dot key which indicates that he wants to enrich this object with information. The system will show the user list of expressions that can add information on this object. In CG terms, the system will fill the list with items which fall in one of the following three categories: + Relations that can be created by the system and their selectional restrictions are such that they allow the modified object as a source for the relation.</Paragraph> <Paragraph position="4"> + Properties that can be added to the concept object such as name and quantity.</Paragraph> <Paragraph position="5"> + Concept types that expect relations, the first of whom can connect to the new concept. For example the concept type &quot;Eat&quot; expects a relation &quot;Agent&quot; and a relation &quot;Patient.&quot; The selectional restriction on the destination of &quot;Agent&quot; will be for example &quot;Animate&quot;. Therefore the concept &quot;Eat&quot; will appear on the list of an object of type &quot;Dog&quot;.</Paragraph> <Paragraph position="6"> The author can modify and add information to the active object by pressing the dot key. An object which itself modifies an object previously entered, can be modified with new relations, properties and concepts in the same manner. The global context is updated whenever a new instance is created in the utterances. When the author has finished composing the utterance, the system will update the local context and will add this information to the generated natural language document.</Paragraph> <Paragraph position="7"> The comma operator (&quot;,&quot;) is used to define sets in extension. For example, in Fig.3, the set &quot;salt and pepper&quot; is created by entering the expression #sa,#pe. The set itself becomes an object in the context and is assigned its own identifier. null The dot notation combined with named variables allows for easy and intuitive editing of the CG data. In addition, the organization of the document as defaults, participants and context (local and global) - provides an intuitive manner to organize documents.</Paragraph> <Paragraph position="8"> Propositions, after they are entered as utterances, can also be named, and therefore can become arguments for further propositions. This provides a natural way to cluster large conceptual graphs into smaller chunks.</Paragraph> <Paragraph position="9"> The text generation component proceeds from this information, according to the following steps: + Pronouns are generated when possible using the local and global context information.</Paragraph> <Paragraph position="10"> + Referring expression are planned using the competing expressions from the context information, excluding and including information and features of the object in the generated text, so the object identity can be resolved by the reader, but without adding unnecessary information.</Paragraph> <Paragraph position="11"> + Aggregation of utterances which share certain features using the aggregation algorithm described in [Shaw, 1995].</Paragraph> <Paragraph position="12"> Consider the example cooking recipe in Fig.3. The author uses the participants section in order to introduce the ingredients needed for this recipe. One of the ingredients is &quot;six large eggs&quot;. The author first chooses an identifier name for the eggs, for example &quot;eg&quot;. From the initial list of concepts types proposed by the system, we choose the concept type &quot;egg&quot;. Pressing the dot key will indicate we want to provide the system with further information about the newly created object. We choose &quot;quantity&quot; from a given list by typing &quot;qu&quot;. seeing that the word &quot;quantity&quot; was automatically marked in the list. Pressing the space key will automatically open brackets, which indicates we have to provide the system with an argument. A tool tip text will pop to explain the user what is the function of the required argument. After entering number, we will hit the space bar to indicate we have no more information to supply about the &quot;quantity&quot;; the brackets will be automatically closed. After the system has been told no more modification will be made on the quantity, the &quot;egg&quot; object is back to be the active one. The system marks the active object in any given time by underline the related word in the input text.</Paragraph> <Paragraph position="13"> Pressing the dot will pop the list box with the possible modifications for the object. We will now choose &quot;attribute&quot;. Again the system will open brackets, and a list of possible concepts will appear. The current active node in the graph is &quot;attribute&quot;. Among the possible concepts we will choose the &quot;big&quot; concept, and continue by clicking the enter key (the lexical chooser will map the &quot;big&quot; concept to the collocation &quot;large&quot; appropriate for &quot;eggs&quot;). A new folder in the global context view will be added with the title of &quot;egg&quot; and will contain the new instance with its identifier and description as a CG in linear form.</Paragraph> <Paragraph position="14"> Each time a dot or an identifier is entered, the system converts the current expression to a CG, maps the CG to a FUF Functional Description which serves as input to the lexical chooser; lexical choice and syntactic realization is performed, and feedback is provided in both English and Hebrew.</Paragraph> <Paragraph position="15"> The same generated sentence is shown without context (in the left part of the screen), and in context (after reference planning and aggregation).</Paragraph> <Paragraph position="16"> When generating utterances, the author can refer to an object from the context by clicking on the context view. This enters the corresponding identifier in the utterance graph.</Paragraph> </Section> </Section> class="xml-element"></Paper>