| { | |
| "paper_id": "J95-4005", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T02:46:28.295467Z" | |
| }, | |
| "title": "Developing a Nonsymbolic Phonetic Notation for Speech Synthesis", | |
| "authors": [ | |
| { | |
| "first": "Andrew", | |
| "middle": [], | |
| "last": "Cohew", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "University of Reading", | |
| "location": {} | |
| }, | |
| "email": "[email protected]" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "The goal of the research presented here is to apply unsupervised neural network learning methods to some of the lower-level problems in speech synthesis currently performed by rule-based systems. The latter tend to be strongly influenced by notations developed by linguists (see figure 1 in Klatt (1987)), which were primarily devised to deal with written rather than spoken language. In general terms, what is needed in phonetics is a notation that captures information about ratios rather than absolute values, as is typically seen in biological systems. The notations derived here are based on an ordered pattern space that can be dealt with more easily by neural networks, and by systems involving a neural and symbolic component. Hence, the approach described here might also be useful in the design of a hybrid neural~symbolic system to operate in the speech synthesis domain.", | |
| "pdf_parse": { | |
| "paper_id": "J95-4005", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "The goal of the research presented here is to apply unsupervised neural network learning methods to some of the lower-level problems in speech synthesis currently performed by rule-based systems. The latter tend to be strongly influenced by notations developed by linguists (see figure 1 in Klatt (1987)), which were primarily devised to deal with written rather than spoken language. In general terms, what is needed in phonetics is a notation that captures information about ratios rather than absolute values, as is typically seen in biological systems. The notations derived here are based on an ordered pattern space that can be dealt with more easily by neural networks, and by systems involving a neural and symbolic component. Hence, the approach described here might also be useful in the design of a hybrid neural~symbolic system to operate in the speech synthesis domain.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Phonological and phonetic notations have been developed by linguists primarily as descriptive tools, using rewrite-rules operating on highly abstracted basic units derived from articulatory phonetics. Even some connectionist work has followed this tradition (Touretzky et al. 1990 ). The primary aim of these notations is explanation and understanding, and there are difficulties in incorporating them into systems with a practical aim such as deriving speech from text, which tend to be data-driven. One recent study claimed that introduction of linguistic knowledge degrades performance in grapheme-phoneme conversion (van den Bosch and Daelemans 1993). However, typical purely data-driven systems are opaque from a phonetic or phonological point of view. In order to handle many of the very hard problems remaining in speech synthesis, there is a need to develop a basic underlying notation (or method of deriving a notation) that can be parameterized for different speakers. This notation could be based on articulatory phonetics (where a higher-level task, such as grapheme-phoneme conversion, is being performed) or on a spectral/perceptual measure of similarity, for more low-level tasks such as duration adjustment. This notation would ideally be represented in a low-dimensional, topological space so as to be both perspicuous and flexible enough to use in further nonsymbolic modules.", | |
| "cite_spans": [ | |
| { | |
| "start": 258, | |
| "end": 280, | |
| "text": "(Touretzky et al. 1990", | |
| "ref_id": "BIBREF15" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Background and phonetic motivation", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "Existing synthesis-by-rule (SBR) systems (Allen et al. 1987) have been concerned with text-to-speech conversion, and have made use of a segmental approach derived from traditional phonology. Among the simplifying assumptions remaining from this approach are that transitions into and out of a consonant are identical, and that the same transition may be used in each CV combination, regardless of the larger phonetic environment. These assumptions need to be modified in a principled manner rather than by tables of exceptions.", | |
| "cite_spans": [ | |
| { | |
| "start": 41, | |
| "end": 60, | |
| "text": "(Allen et al. 1987)", | |
| "ref_id": "BIBREF0" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Background and phonetic motivation", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "It has been argued by phoneticians that articulatory models cannot account for all the variability found in natural speech (Bladon and A1-Bamerni 1976; Kelly and Local 1986) . Therefore, there is a need to find ways of incorporating other sources of variability into synthetic speech, including, for example, the feedback a talker receives from the perception of their own voice. Evidence that such feedback affects speech is the degradation seen in the speech of persons with acquired deafness. One possible way to introduce this kind of variability is through the development of representations that encode (in a reduced dimensionality) a range of examples of the phenomenon to be accounted for. Formant data can be used to introduce a perceptual measure of similarity (see section 3 below).", | |
| "cite_spans": [ | |
| { | |
| "start": 123, | |
| "end": 151, | |
| "text": "(Bladon and A1-Bamerni 1976;", | |
| "ref_id": "BIBREF1" | |
| }, | |
| { | |
| "start": 152, | |
| "end": 173, | |
| "text": "Kelly and Local 1986)", | |
| "ref_id": "BIBREF6" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Background and phonetic motivation", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "This report describes the theoretical motivations of an experimental system that has been implemented as a set of shell scripts and 'C' programs; not all of the technical details of this system have been finalized, and it has not been formally tested. While formants have been made use of as training data (as well as acoustic tube data), as yet no use has been made of a formant synthesizer for creating the output speech, due to the need for handcrafting of values. At present, waveform segment concatenation is being used to explore a parametric duration model based on the kind of proximitybased notations described here.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Background and phonetic motivation", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "In outline, the Self-Organizing Map (SOM, Kohonen 1988) approximates to the probability density function of the input pattern space, by representing the N-dimensional pattern vectors on a 2D array of reference vectors in such a way that the resulting clusterings conform to an elastic surface, where neighboring units share similar reference vectors. This algorithm and Learning Vector Quantization (LVQ) are described in Kohonen (1990) , which has practical advice for implementation, and in more theoretical detail in Kohonen (1989) .", | |
| "cite_spans": [ | |
| { | |
| "start": 422, | |
| "end": 436, | |
| "text": "Kohonen (1990)", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 520, | |
| "end": 534, | |
| "text": "Kohonen (1989)", | |
| "ref_id": "BIBREF11" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "It has been widely noted that 2D representations of speech are useful where there is a need to transmit information to humans at a phonetic level--for example, in tactile listening systems (Ellis and Robinson 1993) . If a speech synthesis system has a phonetic interface or level of operation, it is then possible to introduce learning techniques for subsequent modules (e.g., those which calculate durations or an intonation contour) and to have an idea of what is happening, in phonetic terms, when things go wrong, and therefore how the training program or learning method may be adjusted. There is a long tradition of two-dimensional representations of formant data in attempts to classify vowels, going back at least to the study of Peterson and Barney (1952) . Another type of advantage lies in the flexibility given by the very large dimensionality reductions achievable by Kohonen's technique. These reductions are possible even where the input pattern space may be only sparsely populated, yielding a flexible encoding with not too many degrees of freedom. It is possible for Kohonen's technique to work in 3D (3D maps have been produced by the author, but are more difficult to work with and are still undergoing evaluation). In 4D or above, interpretation becomes much more difficult. Refinements such as the Growing Cells technique (Fritzke 1993) might be preferable to a move to higher dimensionality, so as to retain transparency of the notation and a possible link to symbol-based stages of operation. Figure 1 shows a map resulting from applying the SOM algorithm to phoneme feature data. The following nine binary articulatory features were used: continuant, voiced, nasal, strident, grave, compact, vowel height(I), vowel height(2), and round. The features hl and h2 are used for height simply because there are three possibilities: k open, mid and closed, which cannot be encoded by a binary bit} In this case, the point is not to do feature extraction (since the features are already known), but to provide a statistical clustering in 2D that can indicate whether the features chosen provide a good basis for analysis. Figure 1 suggests that phoneticians have 'got it right' in that the features do result in a clustering of similar sounds such as stops, fricatives and nasals, as well as the more obvious separation between vowels and consonants. It is worth pointing out that neither the SOM nor the LVQ algorithm handles raw data (such as waveform values or image intensity values), but each operates on data such as spectral components or LPC coefficients that are themselves the output of a significant processing stage, and can justifiably be called features. The phoneme map is produced by a single Kohonen layer that self-organizes using the standard algorithm (Kohonen, 1990) , taking as input nine articulatory features commonly used by phoneticians to describe the possible speech sounds. The features were designed so that any phoneme (or syllable) may be uniquely specified as a cluster of features, without reference to specific units (segments such as phones, syllables, etc.)--any feature may run across unit boundaries. Figure I shows a 12 x 8 map created (as are all the following maps) with hexagonal connections in the lattice indicating which units are neighbors. A monotonically shrinking 'bubble' neighborhood was used in all the maps shown here. Kohonen refers to this type of kernel as a bubble because it relates to certain kinds of activity bubbles in laterally connected networks (see Kohonen 1989) . relevant information is captured in the formant trajectories. Maps based on acoustic tube data computed from the LPC coefficients have also been created, with much the same kind of results as seen in the formant maps. That the results should be similar is to be expected as this data is essentially spectral, and bears little resemblance to real vocal tract data. Experiments are currently being carried out to determine whether these maps or those based on formants will work better as part of h prototype speech synthesis system. To factor out the influence of the initial configuration of the network (the reference vectors are initialized to small random values), twenty trials were run on each data set, and the map with the lowest quantization error (QE) was selected as the best. The QE is simply the mean error over the N pattern vectors in the set,", | |
| "cite_spans": [ | |
| { | |
| "start": 189, | |
| "end": 214, | |
| "text": "(Ellis and Robinson 1993)", | |
| "ref_id": "BIBREF3" | |
| }, | |
| { | |
| "start": 738, | |
| "end": 764, | |
| "text": "Peterson and Barney (1952)", | |
| "ref_id": "BIBREF13" | |
| }, | |
| { | |
| "start": 2789, | |
| "end": 2804, | |
| "text": "(Kohonen, 1990)", | |
| "ref_id": "BIBREF12" | |
| }, | |
| { | |
| "start": 3533, | |
| "end": 3546, | |
| "text": "Kohonen 1989)", | |
| "ref_id": "BIBREF11" | |
| } | |
| ], | |
| "ref_spans": [ | |
| { | |
| "start": 1517, | |
| "end": 1525, | |
| "text": "Figure 1", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 2139, | |
| "end": 2147, | |
| "text": "Figure 1", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 3157, | |
| "end": 3165, | |
| "text": "Figure I", | |
| "ref_id": null | |
| } | |
| ], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "Y~zt=l IIx(t) -me(011 QE N", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "where x(t) is the input vector and mc the best matching reference vector for x(t).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "In order to compare QEs, the topology (form of lateral connections) and adaptation functions must be the same, since the amount of lateral interaction determines the self-organizing power of the network. In the simplest case of competitive learning the neighborhood contains only one unit, so a minimal QE may be achieved, but in this case there is no self-organizing effect.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "Schematically, then, resynthesis would take place on the basis of a trajectory across a diphone map. The trajectory could be stored simply as a vector of co-ordinates that are 'lit up' on the map. These vectors would occupy little storage space, and might be passed as input to a further SOM layer to try to cluster similar sounding words. The time-varying, sequential properties of speech, which are difficult for neural nets to handle, can thus be modeled as a spatial pattern in an accessible and straightforward manner. Vectors of addresses would be completely different (e.g., the endpoints Sym Example \"air\" /air/ \"all\" /bard/ \"e\" /bed/ \"oa\" /oak/ \"ie\" /pie/ \"oi\" /oil/ \"oo\" /good/ \"ai\" /pain/ \"oor\" /poor/ \"er\" /bird/ \"aw\" /board/ \"i\" /bid/ \"uu\" /brood/ \"ee\" /bead/ \"u\" /bud/ \"o\" /body/ \"uh\" /above/ \"ou\" /out/ \"eer\" /ear/ \"dr\" /art/ ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Application of the SOM to phoneme data", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "Clustering of data with Sammon's mapping.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Figure 2d", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "concatenation procedure, on which various enhancements based on the SOM are being tried, which will be more fully described in future reports. Using the examples given on a record supplied with Klatt's (1987) review article, informal comparison shows a high degree of variability in quality of the sentences generated: the best are comparable with the diphone concatenation methods (which have better transitions than DECtalk, even if the prosody is in some cases not as well developed), while the worst are highly unnatural, but usually intelligible.", | |
| "cite_spans": [ | |
| { | |
| "start": 194, | |
| "end": 208, | |
| "text": "Klatt's (1987)", | |
| "ref_id": "BIBREF9" | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Figure 2d", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The outline of a conventional SBR system has a series of symbolic stages, assuming a modularity of data at each level, before the final low-level stage ('synthesis routines') calculates the synthesizer parameters. The essential feature is the 'abstract linguistic description', which must be derived before any attempt is made to calculate parameter values. In the proposed system, this middle stage is replaced by the SOM stage, which introduces a learned notation based on acoustic data. Generation of an intonation contour, though this has been implemented with neural nets, is probably best handled with rules as it is almost purely a prosodic (i.e., sentence level) matter. The SOM coding replaces the linguistic description, and leads to direct access of waveform values for a given diphone, which then become default values for the next stage to operate on. In conclusion, arguments have been presented for the use of nonsymbolic codings as the central stage of a text to-speech system. These codings are both closer to the acoustic domain and capable of greater flexibility than the standard phonetic notations. Additional sources of variability, such as stress and emotional quality, could also be accounted for with this kind of trajectory in a lowdimensional space, rather than attempting to derive a speaker-independent symbolic notation. These maps are also capable of being operated on by a neural network in further processing stages, opening the way to a different type of phonetics based on a multitude of soft constraints rather than the rigid phoneme and rewrite rule.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and further work", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Further work is needed to investigate the usefulness of the SOMs in speech synthesis, and how they may be integrated in a hybrid system that uses rule-based prosody. Other data sets need to be explored to introduce other kinds of variability. It would also be important to determine whether the distance measure provided by the diphone maps correlates better with subjective perception of the mismatch between successive diphones than more standard measures of spectral distance, such as various distance measures between frames of cepstral coefficients.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusion and further work", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "Thanks to John Local for providing the basis for the data.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "Thanks to Stephen Isard of Edinburgh CSTR and Linda Shockey of the Linguistics Department, Reading University for help with diphones and related matters. The author is grateful to John Local for enthusiasm and help with phonetics. Any remaining mistakes are the author's responsibility. Thanks to the Laboratory of Computer and Information Science, Helsinki University of Technology for making available their 'SOM3~AK ' software, used here with minor modifications.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Acknowledgments", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "From text to speech: The MITalk system", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Allen", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "M", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Hunnicutt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Klatt", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Allen, J., Hunnicutt, M. S., and Klatt, D. H. (1987). From text to speech: The MITalk system. Cambridge University Press.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Data-Oriented Methods for Grapheme-to-Phoneme Conversion", | |
| "authors": [ | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Bladon", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Bamerni", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "", | |
| "middle": [], | |
| "last": "Van Den", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [], | |
| "last": "Bosch", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "W", | |
| "middle": [], | |
| "last": "Daelemans", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1976, | |
| "venue": "Proceedings, 6th Conference of the European Chapter of the ACL", | |
| "volume": "4", | |
| "issue": "", | |
| "pages": "137--150", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Bladon, A., and A1-Bamerni, A. (1976). \"Coarticulation resistance in English.\" Journal of Phonetics, 4, 137-150. van den Bosch, A., and Daelemans, W. (1993). \"Data-Oriented Methods for Grapheme-to-Phoneme Conversion.\" Proceedings, 6th Conference of the European Chapter of the ACL, Utrecht, April 1993.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Generative and Non-Linear Phonology", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Durand", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Durand, J. (1990). Generative and Non-Linear Phonology. Longman, London.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "Growing Cell Structures--A Self-organizing Network for Unsupervised and supervised Learning", | |
| "authors": [ | |
| { | |
| "first": "E", | |
| "middle": [ | |
| "M" | |
| ], | |
| "last": "Ellis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Robinson", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "A", | |
| "middle": [ | |
| "J" | |
| ], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1993, | |
| "venue": "International Computer Science Institute Technical Report", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Ellis, E. M., and Robinson, A. J. (1993). \"A Phonetic Tactile Speech Listening System.\" Cambridge University Engineering Department Technical Report, CUED/F-INFENG/TR122, May. Fritzke, B. (1993). \"Growing Cell Structures--A Self-organizing Network for Unsupervised and supervised Learning.\" International Computer Science Institute Technical Report TR-93-026, May.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Autosegmental and Metrical Phonology", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Goldsmith", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Goldsmith, J. (1990). Autosegmental and Metrical Phonology. Blackwell, Oxford.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Modelling Human Vowel Identification Using Aspects of Formant Trajectory and Context", | |
| "authors": [ | |
| { | |
| "first": "C", | |
| "middle": [ | |
| "B" | |
| ], | |
| "last": "Huang", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "Speech Perception, Production and Linguistic Structure", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Huang, C. B. (1990). \"Modelling Human Vowel Identification Using Aspects of Formant Trajectory and Context.\" In Speech Perception, Production and Linguistic Structure, edited by Y. Tohkura, E. Vatikiotis-Bateson, and Y. Sagisaka. Proceedings, ATR workshop, Kyoto, Japan, November 1990, IOS press, Oxford, UK.", | |
| "links": null | |
| }, | |
| "BIBREF6": { | |
| "ref_id": "b6", | |
| "title": "Long-domain resonance patterns in English", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Kelly", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Local", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1986, | |
| "venue": "Proceedings IEEE Speech Input~Output Conference", | |
| "volume": "258", | |
| "issue": "", | |
| "pages": "77--82", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kelly, J., and Local, J. (1986). \"Long-domain resonance patterns in English.\" In Proceedings IEEE Speech Input~Output Conference, Pub. No. 258, 77-82.", | |
| "links": null | |
| }, | |
| "BIBREF7": { | |
| "ref_id": "b7", | |
| "title": "Speech processing strategies based on auditory models", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Klatt", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1982, | |
| "venue": "The Representation of Speech in the Peripheral Auditory System", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Klatt, D. H. (1982a). \"Speech processing strategies based on auditory models.\" In The Representation of Speech in the Peripheral Auditory System, edited by R. Carlson and B. Granstrom, Elsevier, Amsterdam.", | |
| "links": null | |
| }, | |
| "BIBREF8": { | |
| "ref_id": "b8", | |
| "title": "Prediction of perceived phonetic distance from critical band spectra: a first step", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Klatt", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1982, | |
| "venue": "Proceedings IEEE ICASSP-82", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "1278--1281", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Klatt, D. H. (1982b). \"Prediction of perceived phonetic distance from critical band spectra: a first step.\" In Proceedings IEEE ICASSP-82, 1278-1281.", | |
| "links": null | |
| }, | |
| "BIBREF9": { | |
| "ref_id": "b9", | |
| "title": "Review of text-to-speech conversion for English", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "H" | |
| ], | |
| "last": "Klatt", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1987, | |
| "venue": "JASA", | |
| "volume": "82", | |
| "issue": "3", | |
| "pages": "737--793", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Klatt, D. H. (1987). \"Review of text-to-speech conversion for English.\" JASA 82(3), 737-793.", | |
| "links": null | |
| }, | |
| "BIBREF10": { | |
| "ref_id": "b10", | |
| "title": "The 'neural' phonetic typewriter", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kohonen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1988, | |
| "venue": "IEEE Computer", | |
| "volume": "21", | |
| "issue": "", | |
| "pages": "11--22", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kohonen, T. (1988). \"The 'neural' phonetic typewriter.\" IEEE Computer 21, 11-22.", | |
| "links": null | |
| }, | |
| "BIBREF11": { | |
| "ref_id": "b11", | |
| "title": "Self-Organization and Associative Memory", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kohonen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kohonen, T. (1989). Self-Organization and Associative Memory. Springer Verlag, 3rd ed.", | |
| "links": null | |
| }, | |
| "BIBREF12": { | |
| "ref_id": "b12", | |
| "title": "The Self-Organizing Map", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Kohonen", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "IEEE Proceedings", | |
| "volume": "78", | |
| "issue": "9", | |
| "pages": "1464--1480", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Kohonen, T. (1990). \"The Self-Organizing Map.\" IEEE Proceedings 78(9), 1464-1480.", | |
| "links": null | |
| }, | |
| "BIBREF13": { | |
| "ref_id": "b13", | |
| "title": "Control methods used in a study of the vowels", | |
| "authors": [ | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Peterson", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "H", | |
| "middle": [], | |
| "last": "Barney", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1952, | |
| "venue": "ASA", | |
| "volume": "24", | |
| "issue": "", | |
| "pages": "175--184", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Peterson, G., and Barney, H. (1952). \"Control methods used in a study of the vowels.\" ]ASA 24, 175-184.", | |
| "links": null | |
| }, | |
| "BIBREF14": { | |
| "ref_id": "b14", | |
| "title": "A nonlinear mapping for data structure analysis", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Sammon", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1969, | |
| "venue": "IEEE Trans. Computers, C", | |
| "volume": "18", | |
| "issue": "", | |
| "pages": "401--409", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Sammon, J. W. (1969). \"A nonlinear mapping for data structure analysis.\" IEEE Trans. Computers, C-18, 401-409.", | |
| "links": null | |
| }, | |
| "BIBREF15": { | |
| "ref_id": "b15", | |
| "title": "Rules and Maps III: Further Progress in Connectionist Phonology", | |
| "authors": [ | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "S" | |
| ], | |
| "last": "Touretzky", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "D", | |
| "middle": [ | |
| "W" | |
| ], | |
| "last": "Wheeler", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "Iii", | |
| "middle": [], | |
| "last": "Elvgren", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1990, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Touretzky, D. S., Wheeler, D. W., and Elvgren III, G. (1990). \"Rules and Maps III: Further Progress in Connectionist Phonology,\" School Of Computer Science, Carnegie Mellon, Technical Report CMU-CS-90-138", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "text": "Figure 1 Clustering of phoneme data (8 x 12).", | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF1": { | |
| "uris": null, | |
| "text": "Figure 2a", | |
| "type_str": "figure", | |
| "num": null | |
| }, | |
| "FIGREF2": { | |
| "uris": null, | |
| "text": "Figure 2c Clustering of data for transitions into and out of 'p'.", | |
| "type_str": "figure", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |