Semantic Fingerprint & Bearing Words
Intrinsic Semantic
Intrinsic semantics according to ai-one™, Walter Diggelmann, refers to the meaning of a text before it is read. Let's assume that a text is completely unknown to us, without any information about the author or origin and without any reference to other texts. Nevertheless, it has its own meaning, which arises from the grammar and style - we call this intrinsic or inherent semantics.
Hans-Georg Gadamer argues that every reader brings a pre-understanding with them - shaped by knowledge and experience. While reading, this gives rise to an initial understanding that broadens the horizon and at the same time changes the pre-understanding. With each further engagement, a deeper or new understanding develops. This process can continue indefinitely.
To translate a text correctly, one must understand it - but to understand it, one must first translate it. This tension is what Gadamer calls the hermeneutic circle: the whole is only revealed through the parts, and the parts only through the whole.
The paradox:
What is to be understood must already somehow be understood beforehand.
Semantic Fingerprint
The semantic fingerprint, according to ai-one™, Walter Diggelmann, transfers the intrinsic semantics of a text into a multidimensional vector space for storage and automated processing. Deep learning creates a digital twin of the text being searched for - and later even of the user himself in relation to changing information and security needs.
Starting from the intrinsic semantics, additional features are defined, normalized and mapped as data points in the vector space - comparable to the papillary lines of a biological fingerprint. In addition to text-internal properties such as frequency, exclusivity, position and networking of key words, contextual factors are also included: time, place, author, motivation, as well as the plausibility and relevance of the statement to reality. The vector space can have up to 100,000 dimensions.
The Fingerprint Similarity Score (FPSS) serves as a means of comparison: a value between 0 (no match) and 1 (complete match). Intermediate values mark the graduated degree of similarity.
Semiotic
Semiotics - also called sign theory - is the science of sign systems of all kinds (e.g. pictorial writing, gestures, formulas, language, traffic signals). It is applied in the humanities, cultural, economic and social sciences.
In text-oriented linguistics, it describes that two messages with the same syntax can have different meanings. Semantics is strongly influenced by time, place and circumstances.
The triangular relationship between syntax (relationships between signs), semantics (meaning of linguistic units) and pragmatics (reference to time, place and users) is therefore essential for interpreting text messages.
Since every message is linked to the place and time of the author, a text without context or intrinsic semantics is difficult to understand.
Semantic Matcher
Creating new texts, rewriting texts or explaining content, describing ideas and much more, are all functions that modern chat robots can do best.
But there is another essential function that is extremely important in the automation of processes, and that is the semantic match!
This is the function of comparing two text sections (paragraphs) in terms of content and outputting a value as to how exactly these texts semantically match. So the same statements, even if different words and stylistics were used.
The semantic matcher uses the semantic fingerprint, which has made texts machine-readable, and compares and evaluates these texts. There are customers who use this technique when comparing and evaluating twenty thousand texts against millions of texts. Such matches sometimes take less than a minute!
