Minimalist Grammars and Natural Language Processing (LMG) - Giving meaning to the language

• The feature lists of Minimalist grammars represent a good model of the syntax of natural language
• The lambda calculus can be used to represent the semantics of natural language
• The combination of minimalist grammar and lambda calculus can be constructed in such a way, that it can be used to accurately parse and generate natural language with computer systems in real time

A NLG was implemented for basic lexicons that automatically selects words according to predicate logical expressions and constructs grammatical correct sentences that can be understood according to the initial expression. Additional a left corner parser inspired by Stanojevic [Stan18], was modified to construct a semantic meaning, while parsing sentences with the help of the feature list of lexicon according to minimalist grammars.

Intelligent Information Processing (scientific accompanying research)

The goal of this project is related to modeling and optimization of continuous industrial processes through mathematical methods

Quantum-Inspired Behavior Controller for a Remake of C. E. Shannon's Electrically Controlled Mouse

• Refine and enhance Ingo Schmitt's Hilbert space representation theory of relational databases, e.g., handling of uncertain knowledge or NIL values,
• Devise a representation theory of semantic states, expressed by feature-value relations, in Fock space,
• Devise quantum-inspired algorithms, i.e., Hilbert and/or Fock space operator compositions, for agent actions, attention, natural language communication, ontology inference, and coping,
• Construct a test bed and demonstrator for the application example of Shannon's mouse-maze system, and
• Ultimately: devise quantum computer algorithms for the behavior control of cognitive agents.

• First steps taken towards representation theories and modeling of the behavior controller (see bibliography below),
• First version of Matlab toolbox called FockBox published.

(Click or tap on images to enlarge)

• Project bibliography [-- UBICO, keywords="btuktqtheseus" --]
• Project page @ResearchGate

• [-- RTHESEUS --]
• Behavior Control of Intelligent Machines (VIM)

Maschineller Spracherwerb. Verstärkungslernen minimalistischer Grammatiken

We are developing a reinforcement learning algorithm for numeral words of natural languages. Numerals have an obvious connection between syntax and semantics, e.g. "two hundred and fourty-five" means 2 * 100 + 40 + 5. Our algorithm is intended to not only learn words but also to predict their integer meaning.

State-of-the-art algorithms try this by slot-filling of digit, i.e. they learn how to decide that 'two' is the hundred, 'four' is the ten… This requires a high degree of individual supervision for single languages – especially if their number systems are not even decimal. Instead, we use more general constraints of numerals word structure that are derived from Hurford's theory of the Packing Strategy [Hur06].
Our algorithm forms new words via exchange of subnumerals and their meaning is predicted by linear regression, see figure.

The exchange of subnumerals works by generalizing expressions like "two hundred and fourty-five" and converting them to an output set of words (X,Y) |--> "X hundred and Y".
In order to decide which subnumerals become generalized and which do not, we use general size and divisibility arguments. We represent the numeral systems by a minimalist grammar (MG).

Moreover, we want to investigate the possibility to utilize search engines as supervisors: Running the wrong word “fiveteen” in a search engine will display significantly less search results opposed to a search of the correct word “sixteen”. Based on that, we attempt to find a model to decide whether or not a word is correct.

• Natural numeral grammars can be efficiently represented as Minimalist Grammars
• Those Minimalist Grammars can be efficiently generated based on a learning algorithm that
  • reinforces itself by linear regression and
  • is supervised by big search engines

Cognitive Systems are adaptive, adaptable and can act target‐oriented even with uncertain and incomplete information. So far Artificial Intelligence (AI) is used as an umbrella term for a diversity of methods for making data available and usable. This project (part of Innovation Hub 13) focuses on the integration of AI‐methods into a cognitive system. Such a system gathers information, reaches decisions and recommends actions. The desired cognitive system requires integration of three key technologies of AI: databases, text processing and behavior control. Initially, we create a methodical classified collection of concepts (thesaurus). Because in this project the raw data is text, the relevant terms and relations get extracted through syntactic and semantic analysis. The resulting structures are then stored in databases for universities, public authorities and industry. On this basis behavior control creates a scored matching and recommends partners and funding.

UBICO keyword="btuktihre"
Innovation Hub13 @Innovative Hochschule

Sprachassistent für Obersorbisch

Upper Sorbian is an extinction-endangered minority language in Upper Lusatia and has around 20,000-25,000 speakers. Correspondingly, it is not used frequently in the internet. The lack of data makes the language hard to access for Big Data technologies like speech recognizers or translation machines. As communication with machines is becoming more and more important for humans, Upper Sorbian needs to work in this communication to keep its usefulness in the future. The challenge of this project is to get away with just little data. We can rarely work with statistical models, so instead we directly implement specific expert knowledge in the machine.

• Improvement of the acustic model by specific training of native Sorbian phonemes
• Introduction and application of a grapheme-phoneme-lexicon
• Improvement of the speech model by implementation of grammar
• Catalog of criteria for voice recordings

• Universal Cognitive User Interface (UCUI)
• Machine Language Acquisition. Reinforcement Learning of Minimalist Grammars (MinGLear)

Verhaltenssteuerung für intelligente Maschinen (VIM)

We want to demonstrate firstly the basic mechanisms of our architecture through the physical implementation of our experiment system and subsequently the coping on the same environment. In addition, we aim to transform the linguistic information into semantic information by expanding our PAC with a communication circle in order to create an integrated system that brings the sensory and linguistic information via minimalist grammar formalism on a common ground.

The Cognitive Agent is realized as a mouse in a maze whose behavioral control takes place in a deterministic as well as in a non-deterministic environment simulation. The tasks  such as finding objects and exploration are implemented on the common basis of Markov Decision Processes (MDP) efficiently. The structured data on the semantic databank and the interaction with the environment lead to an inner stage being created during the exploration phase, on which problem-solving programs are carried out risk-free and thus contributes to damage avoidance and goal assessment.

Project bibliography [-- UBICO, keywords="btuktvim" --]

C. E. Shannon: Theseus - An Electrically Controlled Mouse (Video)

Recent speech dialog systems and cognitive user interfaces allow natural verbal human-machine-interaction and achieve an excellent performance. However, leading commercial solutions heavily rely on transmitting sensitive user information (personal data, voice recordings, etc.) through public networks and on processing, storing and analyzing these data on servers of service providers.

The goal of the UCUI project is to develop a cognitive user interface for intuitive interaction with arbitrary electronic devices which ensures privacy by design. That means that all information processing is done on the device and that no user data ever leave the interface. To this end we develop a stand-alone hardware module doing all signal, speech and cognitive information processing. The main technical challenge lies in building a small, energy-efficient system which achieves an acceptable performance without relying on external computational power and memory.

Interaction with the UCUI device takes place through speech, acoustic and visual symbols, and gestures. The system shall be capable of learning from the behavior of users in order to improve its function. Multiple modules will be able to cooperate (distributed microphone array, task assignment, etc.) over a strongly encrypted wireless connection. The system design is based on a study of user-machine interactions in a real home-automation scenario and will take into account relevant legal and ethical aspects.

Project Bibliography
UCUI @BMBF
Project Server (internal)

Miniaturisiertes System zur intelligenten Signaverarbeitung

There is an increasing demand for intelligent technical systems on a variety of application fields. For this reason, Sinus Messtechnik GmbH, a manufacturer of measurement devices, intends to develop and market a new hardware module for intelligent signal processing. Such an intelligent system can be a very effective aid for measurement engineers and help to minimize the risk of errors, particularly for complicated and safety-relevant measurement tasks.

In this project we develop a miniaturized, modular, and universal hardware module for intelligent signal processing and pattern recognition. Compared to other products -which are less complex, less universal, or both- our module will be very compact and suitable for integration into off-the-shelf handheld computers. Currently, a custom hardware module with algorithms implemented on FPGAs or ASICs is the only possible technical realization for a miniaturized system of such complexity.

The MiSIS device will not only allow measurements. It will also be capable of an automatic, knowledge-based interpretation of the measurements results. A modular hardware concept and universal algorithms make the system suitable for a broad variety of measurement and testing tasks.

• Project bibliography [-- UBICO, keywords="btuktmisi" --]

• Universal Cognitive User Interface (UCUI)