Support image, text​
With embedding allow you to have the best search result compared to other solutions

Your adjustment has been included and here's the revised text:

Scope:

Develop AI and machine learning models to enhance processes, amplify data analytics capabilities, and extend the use of Value Partners Group AOAI support tools to marketing and finance divisions. The application of AI and machine learning techniques will cover the following areas in the new system:

- Transcript Summarization
- Correspondence Composition
- Case Classification
- Sensitive Data Masking
- Document Completeness Validation
- Document Key Entity Extraction
- Data Analytics and Modeling

We will utilize the following tools and models:

- Azure AutoML to customize and optimize models for specific needs and scenarios.
- Azure Language Service to leverage a suite of natural language processing capabilities, such as text analysis, translation, speech recognition, and synthesis.
- Azure OpenAI gpt-35-turbo-16k to generate natural language text for tasks like transcript summarization, correspondence composition, case classification, and data analytics and modeling. Gpt-35-turbo-16k is a versatile language model adaptable to various domains and tasks with minimal data and supervision.
- Azure Form Recognizer to extract structured data from unstructured documents like invoices, receipts, and forms, as well as to mask sensitive data and validate document completeness.
- Azure Embedding Services with RediSearch for data recommendation and high-accuracy data matching.

Additionally, we need to prepare a suggested volume of sample data for each AI area for model training by VPG during the development stage. We will also provide data handling and indexing service support with 30 types of data indexing categories, such as financial reports, contracts, working hours in CSV format.

We must specify the estimated resources required from VPG to maintain Azure OpenAI and data training with embedding after the system launch, considering factors like update/training frequency, required data volume, needed staff effort, time needed, and related running costs.

We will explore potential uses of AI for the new system and strategize for future AI growth, considering rapid changes in AOAI.

Deliverables:

- AI and machine learning models for the specified areas, meeting accuracy acceptance criteria as confirmed by VPG.
- Integration of models with API services into Value Partners Group AOAI, Marketing, and Finance support tools and related components.
- Guidelines for maintaining the AI and machine learning engine post-system launch.
- A technical proposal detailing suggested volume of sample data, estimated required resources, and potential AI uses for the new system.

Assumptions:

- Availability of necessary data for model development.
- The scope may be adjusted at various stages based on feasibility and utility.

pdf index tools
form detector
table detector
fomular input and per train
prompt adjustment