|
8 | 8 |
|
9 | 9 | # VimAssistant |
10 | 10 | VimAssistant is a AI powered productivity assistant to [VimKit](https://github.com/codefiesta/VimKit) that helps users complete tasks more efficiently. |
| 11 | +<img width="590" alt="assistant" src="https://github.com/user-attachments/assets/ae438be8-b39c-435d-be0c-365443f4fe4e" /> |
11 | 12 |
|
12 | 13 | ## Overview |
13 | 14 | The VimAssistant package processes natural language (spoken or typed) to take action on a VIM model. |
14 | 15 |
|
15 | 16 | This package utilizes Speech Recognition to transcribe spoken text that gets sent to a CoreML LLM/LAM that will attempt to execute an action on the users behalf. |
16 | 17 |
|
17 | | -Some examples of actions include (but not limited to): |
| 18 | +Some examples of categorized actions include (but not limited to): |
18 | 19 |
|
19 | | -* `"Isolate all HVAC components."` |
20 | | -* `"Create a section box around level 1."` |
21 | | -* `"Hide all walls."` |
| 20 | +* **ISOLATE**: "Isolate furniture", "Only show me the network jacks", "Filter everything but the curtain panels" |
| 21 | +* **HIDE**: "Conceal all ceiling systems", "Hide air terminals", "Remove all walls" |
| 22 | +* **QUANTIFY**: "Count all doors", "What are the total number of access doors?", "How many transformers are there?" |
| 23 | + |
| 24 | +### Label Scheme |
| 25 | +The base model was created with the OntoNotes 5.0 NER annotations which includes: |
| 26 | + |
| 27 | +* **PERSON**: Individual names (e.g., Barack Obama). |
| 28 | +* **ORGANIZATION**: Company or institution names (e.g., Apple). |
| 29 | +* **LOCATION**: Geographical places (e.g., Tokyo). |
| 30 | +* **DATE**: Dates (e.g., May 8, 2025). |
| 31 | +* **TIME**: Times (e.g., 10:00 AM). |
| 32 | +* **EVENT**: Names of events (e.g., World Series). |
| 33 | +* **WORK\_OF\_ART**: Names of works of art (e.g., "Hamlet"). |
| 34 | +* **FAC**: Buildings or facilities (e.g., White House). |
| 35 | +* **GPE**: Geo-political entities (e.g., United States). |
| 36 | +* **LANGUAGE**: Names of languages (e.g., English). |
| 37 | +* **LAW**: Legal names (e.g., The Constitution). |
| 38 | +* **NORP**: National/religious/political group (e.g., Democrats). |
| 39 | +* **CARDINAL**: Cardinal numbers (e.g., 1, 2, 3). |
| 40 | + |
| 41 | +The trained model provides Construction NER annotations: |
| 42 | + |
| 43 | +* **CON-BIM-CATG**: BIM Category - a high-level classification for families and elements, grouping them based on their functional type. |
| 44 | +* **CON-BIM-FAML**: BIM Family - a collection of elements that share common properties, behaviors, and physical characteristics. |
| 45 | +* **CON-BIM-TYPE**: BIM Type - a specific instantiation of a family that defines a unique set of parameters, essentially a variation within a family. Think of it as a specific size, material, or configuration of a particular family, such as a 3' x 6' door within a door family. |
| 46 | +* **CON-BIM-INST**: BIM Instance - a single, unique occurrence of a family type placed within a model. |
| 47 | +* **CON-BIM-LEVL**: BIM Level - a horizontal plane used to define the vertical position of elements like walls, floors, and ceilings. |
| 48 | +* **CON-BIM-VIEW**: BIM View - represents a specific way of looking at the model, whether it's a 2D plan, elevation, section, or 3D view. |
| 49 | + |
| 50 | + |
| 51 | + |
| 52 | +| Component | Labels | |
| 53 | +| -------- | ------- | |
| 54 | +| named entities | CARDINAL, DATE, EVENT, FAC, GPE, LANGUAGE, LAW, LOC, MONEY, NORP, ORDINAL, ORG, PERCENT, PERSON, PRODUCT, QUANTITY, TIME, WORK_OF_ART, CON-BIM-CATG, CON-BIM-FAML, CON-BIM-TYPE, CON-BIM-INST, CON-BIM-LEVL, CON-BIM-VIEW | |
| 55 | +| categories | ISOLATE, HIDE, QUANTIFY | |
0 commit comments