Search

This paper describes the development and application of an algorithm which generates non-linear representations of English text. The algorithm uses the results of a syntactic analysis system and a set of rules which prescribe linkages to generate a graph of a sentence. The shape of these graphs corresponds to the syntax of the sentence; the labels correspond to the vocabulary of the sentence and the edge types correspond to case grammar roles. The sentence graphs can then be interconnected at common nodes and analyzed according to common edges. Preliminary experimentation has yielded promising results. It appears that the algorithm produces a representation of English text which could be quite useful in automatic language processing.

SUMMARY Examination of the criteria employed by ALLEN (1963) in his classification of cross-stratified units suggests (a) that an additional criterion, the three-dimensional shape of cross-strata, should be used in describing and recognizing varieties of cross-stratification, (b) that the character of the lower bounding surface of sets should be expressed in descriptive rather than genetic terms, and (c) that in describing the relation between cross-strata and lower bounding surfaces, an additional term, asymptotic, should be used. The recent discussion of the morphological features of cross-stratification by ALLEN (1963) represents a significant advance in our appreciation of the variability of this type of sedimentary structure. The valuable classification of McKEE and WEIR (1953) generated a concern for precise description of cross-stratification which has, in 10 years, made the classification inadequate. Allen's proposed classification, based as it is on observable morphological criteria and utilising a nongenetic terminology, should promote further rapid progress. Allen has discussed the problems that arise in naming varieties of cross-stratification. He very properly avoids terms with genetic implications and his use of Greek letters results in commendably brief names. Almost inevitably, his chosen nomenclature is not readily applicable in the field due to the difficulty of remembering which letter is associated with which assemblage of morphological features. This difficulty will increase as the number of defined varieties increases. It seems probable that workers will prefer for the most part to give names to varieties of cross-stratification only in the laboratory where reference to literature can be made. The basis for recognition of varieties will be detailed field notes and photographs of morphological features. Nor is this undesirable, for it necessitates the making of detailed observations in the field which will lead to the recognition of new varieties. An easily remembered nomenclature too often leads to the suppression of significant differences in the interests of “pigeon-holing” the structure in question. It is therefore important that all significant morphological features should be noted. Consideration of the differences between certain varieties of cross-stratification.

Items have been studied heretofore for their value as elements of particular tests to the neglect of more fundamental research into the multiple potentiality of items. This article proposes a method of grouping items into “synonymies” comprising all of the items which correlate with a given key item. These synonymies can be used for interpretation of the total meaning of the key item: (1) by inspection of the constituent items and (2) by correlational study of obtained single scores of individual persons. The method is illustrated by four items with inter- and intra-correlations, and characteristics of an ideal background reservoir of items are pointed out.

Publisher Summary This chapter focuses on connection theory of children's language and linguistic theory. It discusses the problem of a child's recognition of nonuniversal aspects of deep structure. It is not the case that deep structure is the same for all languages. Grammatical relations are, by hypothesis, universal, but the order of elements is not. It is not the case that simple, active, declarative sentences always reflect deep structure order directly. It is not the case that the most frequent forms in surface structure reflect deep structure. Traditionally, the determination of base word order depends upon analysis of the transformational component. The most important transformation in German is the verb-second transformation which transfers an element from final position to second position in the creation of declarative sentences. It is the broad applicability of this transformation that makes elegantly simple a description of German that uses a verb-final base structure. The subordinate-clause strategy is a powerful mechanism for the discovery of deep structure.

The report is a collection of Working Papers in Speech Recognition on the following topics: Organization of the HEARSAY II speech understanding system; The DRAGON system -- an overview; Parameter-independent machine segmentation and labeling; A new time-domain analysis of fricatives and stop consonants; Sub-lexical levels in the HEARSAY II speech understanding system; Inference and use of simple predictive grammars; Real-time linear predictive coding of speech on the SPS-41 microprogrammed triple-processor system; A 16-bit A-D-A conversion system for high-fidelity audio research.

This paper as I understand it tries to set up four levels of linguistic representation in accordance with semantic interpretation. I shall go directly to p. 424 where an outline is given as Table I.

As part of a research program aimed at determining the parameters influencing the effectiveness of a message retrieval system, a collection of 10, 000 technical abstracts was indexed and retrieval experiments were conducted with them. Since part of the work involved the development and test operation of an associative retrieval system, basic data about the distribution of words and word strings were gathered in preparing the system for test and trial. These statistics were thought to be of possible interest to other workers in the field and are gathered as a series of loosely connected papers in this volume under the following groupings: Characteristics and Indexing of GE Data Base; Comparison of Manual and Machine Selected Vocabularies; Vocabulary Distribution Studies, and Studies of Content Bearing Units in Text.

THE recent communication on this subject by R. Robinson1 has prompted me to record some observations made about two years ago on this subject.

The standard method of identifying a typewriter as having typed a certain document has been to examine the typed impressions on the document for irregularities or identifying characteristics peculiar to that certain typewriter, as distinguished from all other typewriters of the same class. Among the identifying individualities considered are alignment, characters striking off their feet (striking harder on one side of the character than on the other), slant, and damage to the type face. Positive identification requires a sufficient amount of questioned and exemplar type to eliminate the element of chance and establish a definite pattern. In practice, this has not always been easily accomplished. Frequently a very limited amount of typing is available, and additional difficulty arises when the typewriter in question is fairly new and individual identifying characteristics are few.

This paper describes a system of computer-aided diagnosis using an English Electric KDF9 computer linked to a terminal in a busy clinical department. Data from a series of patients were recorded, coded, and entered into the computer, which then performed a Bayesian analysis and displayed diagnostic probabilities in an adaptable format. Experience in this setting suggests that computer diagnosis may be a valuable aid to the clinician.

著者はブドウ酒酵1づ:の支芽汁培jl.S液から昼光を避けた状態で,cthcrを用いて脚臭を脱臭する性質を有する結晶性の物質を得た。本物質はpapcrchro一matOgralnによるRf値,昇華性,P01ymOrPhiSnl)融点,溶解度,アルカリ摘定,元素分析,紫外線吸収,赤外線吸収,旋光性,X線回折などによって検討した結果,元素分析゛値,赤外線吸収は既知のコハク酸と酷似しており,他の点では異なった性状が認められた。そこで本物質はコハク酸の立体巽性体(cis21;慧およびcis型に近いCauchc型)であろうと推定した。

String analysis has developed out of an attempt to carry out syntactic analysis on a computer, just as, some ten years earlier, transformational analysis developed out of the attempt to normalize texts for discourse analysis. The arrangement of syntax for computability, following in part the method presented in ‘From Morpheme to Utterance’ (Language 22 (1946), 161–83; Paper VI of this volume) was based on an effective procedure for finding in each sentence a sequence (in general, broken) of words which was itself a sentence, belonging to a certain set of minimal sentence structures. This minimal sentence was called the center of the given sentence, and its meaning had an important and central relation to the meaning of the given sentence; this relation can be specified independently of the given sentence. The remainder of the sentence consisted of adjunctions to the center or to the adjunctions; an effective procedure was presented for an ordered determining of these adjunctions, and the ordered adjunctions had an interpretation independent of the given sentence. The original version of this analysis, made for the Univac sentence-decomposing program of 1959, is given in Computable Syntactic Analysis (TDAP 15), 1959 (Paper XVI of this volume).1

Curiously, the ACM First Glossary of Programming Terminology has no entry for the word "document." The American College Dictionary defines it as "a written or printed paper furnishing information or evidence, a legal or official paper." Without becoming lexicographers, we might observe that "write" means "to trace or form (characters, letters, words, etc.) on the surface of some material, as with a pen, pencil or other instrument or means; inscribe."

The purpose of this paper is to define a mathematical model for the study of quantitative problems about translations between universal languages and to investigate such problems. The results derived in this paper deal with the efficiency of the translated algorithms, the optimality of translations and the complexity of the translation process between different languages.

This paper presents the results of research into the development of a semantic-based picture recognition technique. The need for such a technique is reflected in the limitations of syntactic picture recognition techniques, which are discussed in detail. The motivation for developing a semantic-based technique arises from a few intuitive notions described in the paper. A paradigm for semantic picture recognition is realized in the development of a general procedure which makes explicit use of contextual information and bases recognition on intensional class descriptions.

Summary.-Two forms of a novel paper-and-pencil encoding task were developed to measure proactive inhibition during learning when given to children and adolescents. The 2 forms of the encoding task were administered individually to 63 students referred for psychological services throughout a school district from a midwestern city. Results indicated that proactive inhibitory processes were present throughout learning and suggested that this encoding task could substitute, in special cases, for the classical verbal presenration of paired-associate materials that typically require individual administration. An attempt was made to make associative learning an intentional and integral component of a paper-and-pencil encoding task that could be easily and efficiently administered to individuals or large groups of Ss. Such a task, if developed, could substitute in special cases for the often tedious paired-associate technique typically requiring individual administration and a great deal of consideration on the part of the investigator as to the list construction, administration, and reliability. METHOD The combination of letters selected for this encoding task were chosen so that the total sum value of each of two paired-associate lists would approximate that of the other in terms of meaningfulness (m) , as described by Schulz and Underwood ( 1960). Stimulusresponse members were selected from tables presented by Schulz and Underwood for paired combinations of letters and so selected to produce two lists of word combinations with similar total sum value for m. Two experimental encoding tasks were developed utilizing these two lists of similarly scaled stimulus-response members. Each list was incorporated into one form of an encoding task (cf. Fig. 1). One task consists of the paired-members of List X and heretofore referred to as Form X of the encoding task. The Y list was incorporated into Form

In this paper [and in Joshi et al. (1972) which is a sequel to this paper] a new style of formal grammar called String Adjunct Grammars (AG) has been studied. The rules in an AG have a character essentially different from the “rewrite rule” in a Phrase Structure Grammar (PSG). Such a study of formal grammars of different styles is of great interest because each style is well suited for characterizing certain aspects of natural language structure but has inherent difficulty in characterizing certain other aspects. Several subclasses of AG's motivated by strong linguistic considerations have been studied. Linguistic relevance of these grammars and other grammars suggested by this study has also been discussed.

In this paper I shall define the concept logical information distinguished from a customary concept (semantic) information, which is called "factual information" in the paper. In accordance with the contemporary theory of semantic information, if i is an analytic sentence, then i conveys no information. It means thatinf (i) = -log m (i) = 0where i is an analytic sentence and inf is the measure of information. According to our intuition, however, it seems that most analytic sentences convey much more information than "A is A", which is a trivial analytic sentence or tautology. I shall give only one example here.1. Man is man. (S1)2. Man is a rational animal. (S2)Let us suppose that the both sentences Si and S2 are analytic. Then S2 has the same intention, and hence conveys the same amounts of factual information, as S1, whereas S2 has the more complicated intensional structure, and hence conveys much larger amounts of logical information, than S1.

For the storage of information for subsequent retrieval of desired items, two stages of analysis are essential. The first is the determination of the subject content of a given article or paper; the second is the selection of certain words, groups of words, or classification headings by which the subject content is to be represented, either directly or by a suitable coding. Some workers still look forward to the day when it will be possible for the whole of a text to be read and ‘understood’ automatically by a machine; the ‘understanding’ process has been envisaged either as a process of selection of terms by the measure of word frequency, or word‐pair frequency (adjacent terms or terms not too far separated in one sentence) in the text, or by some process of automatic linguistic analysis. Such methods appear unsuitable for several reasons: language, as normally used, is a very difficult medium for exact expression (hence the value of mathematics) and few authors write well enough to avoid all ambiguities; a human reader accustomed to the subject can easily overcome any difficulties due to poor grammar, badly expressed arguments, excess brevity or prolixity in writing and even, sometimes, actual errors; a machine can not do so. Furthermore, the content of a paper is rarely of uniform importance throughout, and it is not worth recording, for subsequent retrieval, details which are merely repetitions of matters described earlier and better elsewhere, and not essential to the main purpose of the paper; for example, in a paper on evaporator design, a description of a standard method of analysis, applied to the contents of the evaporator in determining the efficiency of the design, will not be worth indexing; in a search for analytical methods, retrieval of such a paper would hardly be considered pertinent. A human reader, though far from infallible, can usefully make such judgments.

This paper considers a class of information sources consisting of text and pictures. The text is English language text appearing in scientific and technical documents. The picture sources are the largely schematic pictures that occur in the same class of documents. However, the discussion is broadened slightly to include other picture sources. For a tiny fragment of English, the paper shows how the syntactic structure of text may be described, and then goes on to suggest that a similar analysis may be performed on the class of pictures under study. The description of these two kinds of information sources with a single class of descriptive techniques is suggested as an alternative to the synthetic approach in which artificial languages are specified and then learned and used. The major reason for doing syntactical analysis of such sources discussed here is that several information processing operations, amounting to the interpretation of the information sources looked upon as languages, can be done by the technique of syntax direction which uses the results of syntactic analysis to mediate subsequent processes for manipulating the information tokens. The paper concludes with an illustration of an algorithm for matching the sentences given by a simple grammar against the class of simple pictures which these sentences purport to describe.

Situations constantly arise in which it would be helpful to measure readability or understandability of written material. Most of the available techniques for estimating understandability are difficult to apply and interpret. The paper describes a little-known method--cloze--that offers considerable promise in readability measurement. In addition to describing the cloze procedure, the paper provides specific instructions for administering a cloze test and examples from the professional literature of cloze applications. Attractions, as well as limitations, of the technique are given. Finally, two experiments are described in which the cloze procedure was used. In the first, the technique was shown to be useful with samples of readers having extremely different educational backgrounds: adult basic education students and people in graduate school. The second experiment demonstrated the utility of cloze in measuring the readability of material in another language (Vietnamese), which the experimenters could not read. Cloze differentiated two Vietnamese translations of the same English passage, in agreement with an independent criterion. Although the test subjects who read the material were Vietnamese, neither the experimenters nor the staff people who scored the subjects' responses were proficient in that language. (Author)

This paper describes some pilot experiments in a project to recognize selected words spoken by any talker. For this initial work, a population of 10 talkers is used and the 32 test words are spoken in isolation. In the experiments, a word is represented by a set of regularly spaced time samples of the normalized spectrum envelope, and recognition is achieved by comparison of such a set of samples with a library of stored sets. Limitations on computer storage necessitate a compact method of coding a spectrum sample. Two methods have been compared, one of which classifies spectra into a very small number of types, while the other uses a 24-bit representation. The most serious problem encountered in spectrum matching is the well-known lack of synchronism between corresponding spectral events when phonemically identical words are spoken by different talkers. This paper describes an attack on this problem, which relies on using a number of sets of spectrum samples to represent each word in the stored library of words to be recognized. There are a number of ways of using the scores obtained from matching an unknown word with a library of known words. Some of these are described and some results are given both for the case where the unknown word is known to be in the stored library, and for the case where there is no such limitation.

This paper provides an overview of work on statistical formulations and analyses associated with the problem of identifying persons on the basis of spectral energy representations of acoustical utterances. The investigation has been largely empirical and the paper focuses on the statistical techniques and strategies that have been developed in the context of analyzing two sizeable bodies of data. The problems and procedures to be discussed include: (i) data condensation and representation; (ii) efficient and practical criteria for classification and discrimination; and (iii) strategies for automatic identification of talkers in relatively large populations.

Affix grammars are two-level grammars which are similar to van Wijngaarden's two-level grammars used in the definition of Algol 68. Affix grammars are shown by Koster to be equal in power to van Wijngaarden grammars. They are much more suited to parsing than are the latter, however. Koster, the inventor of affix grammars, suggests a top-down scheme for parsing them, based on recursive procedures. This paper presents a bottom-up scheme for parsing them, based on an extension of Floyd Production Language (FPL). Included is an algorithm, similar to that of DeRemer's, for converting a large class of affix grammars into FPL. The paper concludes by discussing briefly the applicabilities of the conversion algorithm and affix grammars in general, and some possible extensions to Koster's definition of affix grammars.

The paper examines a body of empirical data on S's performing the Thurstone Letter Series Completion task, in order to test the theory proposed by the authors in 1963 for explaining behavior on this task. The data confirm the theory in its main aspects, while indicating the need for some minor extensions and modifications. In particular the data show that subjects first discover the periodicity of the letter series, then construct a description of the pattern, and finally use the pattern description to make an extrapolation. Most of the pattern descriptions used by Ss fall within the pattern language defined in the earlier paper.

In some cases disease causes widespread changes in a number of physical and chemical indexes, without any one index being grossly altered. The difficulty inherent in evaluating numerous small changes may obscure the diagnosis, even though it is conclusive. We show that subtle patterns of change may be recognized using marked cards. These consist of a ``patient data card'' on which is typed a number of scales, each representing one of the measurements, and a ``statrule'' card which contains the statistics for these measurements for the diseases of concern. These cards implement procedures often used in studies of numerical diagnosis made with digital computers. The statistical basis of statrules is discussed in the first part of the paper using primarily a geometric point of view. The second part of the paper shows how to use and construct statrules.

The usefulness of a structural analysis of pretest items in designing an instructional system is illustrated by the application of this technique to 100 addition computation problems. 10 structural characteristics of the problems were derived from the more general objectives of the instructional system (micro-objectives) and used as predictors of problem (mini-objectives) difficulty level in a stepwise regression analysis, in a manner similar to that of traditional studies of readability. The authors point out that regression equations derived in this way can be used to plan lessons in which all problems would be of approximately equal difficulty, and thus could have maximally reliable entrance or pretests, and might also be concerned with only one factor or topic. Lessons could be arranged from easiest to most difficult by using the regression equation to predict lesson difficulty. It is thought that the size of the multiple correlation between difficulty level and the structural variables may be an indication of how well the micro- and mini-objectives of the instructional system have been formulated. In designing and building an instructional system, one of the first steps is to define its objectives. Stolurow (1968) distinguishes three stages in the process beginning with the identification of macro-objectives. These are descriptions of behavior at a very general and abstract level, such as the systems developed by Bloom et al. (1956) and Krathwohl et al. (1964). Micro-objectives are sentences containing verbs that describe an observable type of behavior, while mini-objectives are statements representing the performances observed when an individual is responding to a particular frame in an instructional test or program. Mini-objectives are typically represented by items. These may be used in a pretest to determine whether an individual already has the behaviors or competencies the instructional system is supposed to produce. Can pretest data also be used in designing the instructional system? Do they contain information which, with proper analysis, could guide the construction of instructional materials? This paper presents a method of regression analysis of pretest data which determines a minimal set of item structural characteristics, defined at the level of micro-objectives, that predict the differences in the difficulty level of the mini-objectives (items). Similar regression analyses have been done for decades on reading material to determine their readability (Chall, 1958; Klare, 1963). Suppes, Hyman, and Jerman (1966) also predicted error rate and response time for arithmetic problems from their structural characteristics. This paper not only contributes to this literature, but also attempts to demonstrate how the results of this kind of analysis can be used in designing an instructional system.

The topics began with discussion of almost exclusively syntactic analysis and methods. Beginning with context-free phrase-structure languages, we considered limitations thereof to remove generative syntactic ambiguities (Floyd), and extensions thereto to introduce more context-dependence (Rose). As the conference proceeded we ran through a spectrum of considerations in which the expressions in the languages considered were examined less and less as meaningless objects (the formal, or purely syntactic approach, as in the paper by Steel) and required more and more meaningful interpretations. In other words, we became more and more involved with semantic considerations. It is clear, then, that applications of the study of mechanical languages to programming must involve semantic questions; ADD must mean something more than the concatenation of three (not two) characters. The papers beyond Session 1 were therefore discussing the mechanization of semantics, but in only one case did we hear about the formalization (and hence mechanization) of the specification of the semantics of a language (McCarthy).

Copies of the unmarked scripts of ten candidates in the 1970 B.Sc. (Eng.) Part I paper on Fluid Mechanics were marked by eight separate markers each working to the same marking schedule. For the whole paper (maximum 100 marks) the standard deviation of a candidate's marks about his mean mark fell within the range of 3 to 11 marks. If the results of this investigation are typical the standard deviation in the marks aggregated for N equally weighted examinations is 5√ marks and is independent of the candidate's mean mark.