Translation and typing (OCR) of a patent dossier from DE-EN to UK with reproduction of formulas and tables in Word

There is an archive of 18 scanned PDFs (300-600 dpi) of patent application materials in the field of electrochemical energy storage (Li-ion, electrolytes, cathode materials). Some of the documents are in German, some in English, there are handwritten notes, stamps, small print in tables, as well as formulas, Greek symbols and references to figures. It is necessary to prepare the Ukrainian version of the package for internal use by R&D and lawyers, with accurate typing and preservation of the structure.

Technical environment and output data:
- Input: PDF scans, some pages have skew, noise, different contrast, sometimes a two-column set.
- Output: 1) DOCX with translation and reproduced formatting, 2) a separate TXT/CSV with a list of terms and counterparts (DE/EN-UK), 3) a short report on problem areas (illegible fragments, doubtful places, reading options).
- Preferred tools: ABBYY FineReader or similar with DE/EN OCR support, MS Word with autonumbering, Trados/MemoQ or another CAT (if possible), checking terminology with own glossaries.

What exactly needs to be done step by step:
1) Processing of scans for OCR - alignment, background cleaning, contrast correction, separation of two-column pages, setting of recognition languages.
2) OCR and typesetting - get edited text without loss: headings, subheadings, item numbering, markings (a), (b), (c), tables, units of measurement, indices/superscripts, μ, Ω, °C type symbols.
3) Translation taking into account patent stylistics - stable terminology, correct reproduction of formulations in the Claims/Formula sections, consistency of definitions throughout the package.
4) Reproduction of formatting in Word - save the structure of sections, automatic numbering of points, transfer of tables in edited form, insertion of formulas through the Word formula editor, reproduction of links to figures (Figure/Fig./Abb.) without shifting the numbering.
5) Quality control - selectively comparing at least 10% of pages with the original scan, checking the unity of terms and numerical values, final proofreading of the Ukrainian text.

Limitation:
- Do not use machine translation without editing the patent translation level. If MT is used as a draft, the result must be completely brought to a professional level.
- Do not paraphrase the formulas and statements in the claims - only an accurate translation so as not to change the content.
- Confidentiality: materials may not be uploaded to public online OCR/translation services; work offline or in local tools.

Measurable acceptance criteria:
- DOCX without visible OCR artifacts, with editable tables (not as pictures) and formulas that open in the Word formula editor.
- The logic of the numbering of sections and points has been preserved, there are no gaps/duplicates of points after conversion.
- No discrepancies in numbers, ranges, units of measurement and designations compared to the original.
- Glossary of at least 60 terms with consistent use throughout the document.
- List of all illegible or doubtful passages with page/paragraph reference and suggested reading options.

The volume is approximately 95-120 pages, but with a high density of technical text and a large number of tables/formulas. Attention to detail and experience with patent/chemical texts is essential.