ICE-4005 - Meeting 3
1730073600000
Meeting
Attendance: Alan, Bill
Agenda:
- Follow the evolution of the project
Main discussion points:
- Alan experienced difficulties during the building of the project locally which led to the following discussion on LanguageTool forum
- Bill said that this was exactly the kind of things that was expected to happen, and that a problem leading to a code update on the project is some sort of a progress.
New concepts
- Java:
- Hunspell dictionaries:
- the .aff file format
- HAL: Hyperspace Analog Language, how words proximity can be considered as an analog to their semantic proximity (articles: Nasharuddin and her example here)
- weight (in the case, the proximity between two words), and how one may use similar weights patterns in parallel text to find words translations.
- window size, used to define the wieghts
- Tuples as vectors: I understood that vectors can have many more dimentions that defined in the Euclidian space, and that the maths still holds, especially for the following:
- cosines similarity, Euclidian norm, vector normalisation
- See my conversation with mistral as I digested the new concept
- Vector space model: same concepts as HAL, but used more to categorize document instead of analizing the relationship between words