Quick GuideProducts

LingBench IDE

Home
Company
Technology
Formulas
Download
Job opportunities
Quick Guide
Products
Contact

(IDE is for “integrated development environment")

  Figure 1:

A typical view when working with LingBench IDE

 

What is LingBench IDE?

LingBench IDE is an Integrated Development Environment for Linguists and Computer Linguists to Model a Natural Language by describing the morphology, the syntax, the lexical and other aspects of a language in detail. These high quality models (files) can then afterward be used by other programs so that a computer can handle these aspects of the language autonomously . The innovative and intuitive user interface of LingBench IDE™, based on patented technology,  is a very efficient and unsurpassed powerful way to handle almost every aspect of computer linguistics.

See also LingBench IDE Light.

See also FAQ and Plaform&Pricing further on this page

Grammar Designing Features:

  • Backbone is RTN (Recursive Transition Network), which is conceptually compatible with almost every other modern grammar-modeling paradigm, including BNF.

  • But the RTN backbone has been very much further elaborated, first into ATN (Augmented Transition Network), and far beyond, including probability everywhere and complete programs can be associated with any transition. See below.

  • Free choice of word classes, phrases, features and their names and properties, all of which can be changed easily at any time.

  • Phrases are rich entities that can (but don’t need to) have arbitrary ‘variables’ (e.g. linguistic features or booleans) to coordinate between words or wordgroups or for semantic purposes

  • Transitions can have attached conditions and actions acting on features of words and phrases or, if required, complete complex programs.

  • Automatic or manual probability assignment for transitions and inherent handling of these probabilities up to the parse forest. Each surviving parse gets a probability that a together add up to 100% for one input sentence.

  • Intuitive 3D graphic Phrase Definition by mouse clicking and dragging.

  • Virtually no upper limit on the complexity of models.

  Figure 2:

You can choose to edit your grammar using a 3D graphical user interface. In the View on the right, nodes are added by clicking left, transitions are added by dragging from one node to another. In this way you define paths to go from the upper node to the lower node, thus defining the blueprint of a Phrase definition.

Both Views on the left represent separate grammars in their own: the upper one is the morphology grammar, the lower one the (traditional) syntactic grammar. Such grammars contain wordclasses (or morpheme classes), shown as blue balls, and phrases, shown as purple balls. When you click on a Phrase ball, a View of its 'inside' is shown, like the one on the right. You can then go on editing the 'inside' of the particular Phrase. Clicking inside one of the left grammar views allows adding arbitrary new classes and phrases, and much more.

  Figure 3:

You can freely choose which features you want to be known in the grammar, their names, values and further behavior, and in which (word/morpheme)Classes and Phrases they are relevant. 

 

Morphology Modeling Features:

  • Very similar to syntactic grammar design with all its rich possibilities

  • Automatic coordination with the syntactic level

  Figure 4:

The three levels of lexicon that can be in a language model

 

  Figure 5:

In a lexicon, large amounts of information can be present, which makes it difficult to visualize everything in columns. In LingBench IDE, you can easily change which information should actually be displayed in the columns, by selecting, in a dialog box, the items to be shown from a tree that contains all available information, and their repetitions (for instance, at a given moment, 3 columns show the possible wordclasses. If you want more such columns or none of that kind, the view can be changed by tuning the settings in the dialog box on the left).

Lexicon Development Features:

  • three levels of lexica: morphologic, syntactic and multiword

  • lexica can easily be imported from, or merged or enriched with external sources

  • easy semi-automatic additions to lexicon from corpora containing unknown words

  • origin of entries in the lexicon is kept track of, per word or set of words that were imported together

  • Several 'Tasks' can be defined on a lexicon. Each task contains a description of which properties of which type of entries are to be edited. As the task is carried out (in one or more sessions), the readiness state of each such entry is kept track of. A task allows modifying specific properties of these entries very efficiently and quickly. An example of a simple task: filling out the correct gender of all Nouns that begin with “pre” and that originated from a corpus of BBC news texts.

  • Rich and intuitive set of possibilities to see only a 'filtered' selection of a lexicon, in simple 'DOS' style or with complex regular expressions

  Figure 6:

Setting filters on top of the columns in a lexicon view. Remark that filters apply to sibling columns as well.

 

  Figure 7:

Defining one or more simultaneous "Lexical Tasks"

 

Figure 8:

A typical view of a lexical task session. In many cases, just one mouse click per entry property allows to fill out Task specific properties in the most efficient way possible. The lexical task can be carried out in multiple sessions and target a specific readiness level.

 

  Figure 9:

The integrated parser immediately shows the effect of the latest model changes on sentence in the testcorpus. By clicking on parse tree summaries in the parse forest view one or more parse trees can be examined in detail.

Model Testing/ ‘Debugging’ Features:

One of the major advantages is that the state of the language model can be tested on a testtext, a sentence or a separate wordform, all from within the same environment.

  • Graphic syntactic analysis (parse) displayed as a (3D) tree.

  • Easy browsing among the resulting parse trees in the graphic overview of the parse forest

  • For syntactic as well as for morphologic analysis

  • Statistics on parsibility of the sentences in the testtext

  • Easy swapping of test-texts

  • Informative per word annotation of analysis results (See above figure).

Editor properties:

  • Workspace concept: clear overview of the languages and their components

  • Easy resumption where your work stopped last time.

  • Efficient mouse use: Right clicking makes a menu pop up with the commands that are relevant for the place pointed to, so no unnecessary far mouse moves.

  • Undo/Redo facility of multiple editing steps

  • Autosaving of Language models every few minutes or every few editing actions (so if a crash or other severe problem occurs on your computer, no significant amount of work can be lost).

  • Language models can be saved as a whole, or subaspects like separate lexica can be saved or loaded separately. Also, parse forests and trees can be saved.

  • If the user prefers so, many 3D views can be toggled to textual display

  Figure 10:

Toggling between 3D and textual grammar overview, depending on the preference of the user

Languages and compatibilities:

  • Language independent software

  • Highly language universal

  • Unicode version forthcoming

  Figure 11: the LingBench IDE opening screen

FAQ

(version September 2002)

Q: Can I evaluate LingBench IDE for free?

A: Yes: download the Light version to do this. You can evaluate it during four weeks for free.

Q: Can I start developing from my own old lexicon?

A: Yes, if you can export your lexicon to an ASCII file with TAB or similar separators. You can keep your old wordclass names, but you should define them first into the grammar, before importing the lexicon.

Q:  Can I recycle my old grammar / morphologic models?

A: Importing external formats for grammar isn’t available in version 1.1 yet, but you can quickly enter (draw) the schemes of your phrases or BNF rules. You soon will find out that the LingBench IDE approach gives you a feeling of insight and overview on your old grammar, particularly if it was text based.

Q:  It all looks fancy, but is this a serious language modeling system?

A: What you can’t see is an extremely powerful, versatile and innovative battery of new patented technologies behind the screens, for instance the integrated screamingly fast parser. But why should all this be packed in brown paper?

Q: 3D and animation everywhere, but is this functional? Can I work in textual mode?

A: For complex transition schemes defining complex phrases, 2D is not satisfactory as many crossing arrows confuse the viewer. Only 3D and animation together are extremely efficient in letting the user keep overview, which is of paramount importance for designing high quality models.

The 3D view of parse trees is at present not strictly necessary, but in future versions, more information will be shown with the trees, which isn’t evident in 2D. For grammar overview windows, the 3D can be toggled with a textual, list-like view, which often can be handier.

Working completely in a textual mode to design the grammar is in general not efficient and can easily lead to inconsistencies, so it’s against the LingBench IDE philosophy and not supported at present.

Q: LingBench IDE seems all right, but is there support, and what guarantees are there for the future?

A: There is support included in the standard license, including the right to get a next version for free, and a guarantee that all well documented and repeatable problems will be solved in the new version, except the platform/installation dependent ones. Extended support is subject of a separate contract.

LingBench IDE has been conceived in 1996 and the original designers are still in Natlanco. Natlanco, 100% owned by them, has a strategy of long-term stability and is extremely cautious with financial risks.

Q: Is LingBench IDE a stable product?

A: LingBench IDE is a highly complex product, but is surprisingly small as a program. This is because of the quality of code. We did our very best to get LingBench IDE as reliable as possible, but also we have included functionality to safeguard your work regularly, so that in case of a crash, you easily can recover the situation of a few minutes ago. There is also regular (invisible) internal consistency checking.

Q: Is there a manual? Where can I learn to use LingBench IDE?

A: There is an extensive manual. No courses are given yet, but this can be discussed.

Q: When will the Unicode version be available? Will that system be compatible with Chinese/Japanese/Arabic?

A: The Unicode version just missed the LangTech2002 conference.

Chinese and Japanese are compatible with the present version if Pinyin or another ASCII transcription system is used. Compatibility with Arabic can be expected in Q4 of 2002.

Q: What can I do with the language models made with LingBench IDE?

A:  Natlanco also sells its LingBench SDK product, a software development kit, which contains a software module that can ‘understand’ the format of a Language Model made with LingBench IDE. SDK stands for “software development kit”. Typically, clients for this product develop end application software, like machine translators, search engine systems, grammar/spelling control and correction etc… (see the Natlanco site, “Quick Guide”), and use this (extremely fast) module to do the linguistic analysis of their incoming sentences.

The current SDK is for Windows2000 and XP only. Compatibility with other platforms can be developed on demand.

Q: Does Natlanco itself also sell Language Models? Which ones?

A: Natlanco has several language models, but currently only a model for English is commercialized.

Q: To which domains is LingBench IDE relevant?

A: LingBench IDE is relevant for people and companies who are looking for: (keyphrases)

  • Natural Language Processing Software (NLP Software, NLP Tools)

  • Language Modeling Software

  • Computer Linguistics Software

  • Linguistic analysis software

  • Grammar designing tool

  • Grammar modeling tool

  • Natural language engineering software

  • Natural language engineering tools.

  Figure 11: 

With LingBench IDE there is virtually no upper limit to the quality and complexity of the language model. And the innovative user interface allows unsurpassed levels of efficiency and productivity. The language models can either be exported to your own format or, much better, be used by the engines in the LingBench SDK package, which you can build in in your own application.

Platform &Requirements:

  • PC, minimally 1 GHz, 256 MB RAM, 512MB recommended

  • Windows2000 or XP

  • Screen resolution: min 1024x768, 16 bit colors (high color)

User License:

  • One installation on one PC (academic license: on 3 PCs)

  • Unlimited use (in time) after key code exchanged

  • Support through e-mail to solve fully documented repeatable problems

  • Right to get next version of same product for free, including the fixes of all documented repeatable problems (except installation or computer dependent problems) you correctly reported within the first 2 months after license purchase.

  • Manual + next update of manual  (update by e-mail)

Pricing

  • Professional Version 1.1: € 2499 per licensed installation

  • Professional Version 1.1 for Academic use: € 2499 for 3 licensed installations

  • Extended Service for Professional Version 1.1, during first 3 months: € 1000

  • Light Version 1.2: downloadable shareware, 4 weeks free, then € 199 per licensed installation (download available from May 1, 2003 onward)

|  Home  |   Company   |   Technology   |   Formulas   |   Download   |   Job opportunities   |   Contact   |

© Copyright Natlanco nv 2001

Webdesign
Twixel