Overview
Domain 7 is a tour of the pretrained OCI AI services that developers call through REST APIs, SDKs or the console. OCI Language analyses text: it detects language, finds sentiment at aspect and sentence level, extracts named entities and key phrases, classifies text, detects personal data and translates. OCI Vision analyses images, with image classification, object detection, text detection and face detection, and lets you train custom models on your own labelled images.
OCI Document Understanding extracts text, tables and key-value fields from documents and classifies document types, with pretrained models for receipts, invoices and identity documents. OCI Speech converts audio to text with timestamps and confidence scores and also offers text-to-speech. Oracle's objective for this section is a single item, exploring the services and their APIs; the four topics below split it by service.
How it's tested
Most questions give an input, such as a recorded call, a scanned invoice, a product photo or a batch of reviews, plus a desired output, and ask which service or feature delivers it. Others name a feature and ask which service offers it. Focus on input format first, then on the precise output requested.
-
Choose the service
An input type and goal are given, and you pick Language, Vision, Document Understanding or Speech.
-
Name the feature
You identify the capability that fits, such as named entity recognition, key-value extraction, object detection or transcription.
-
Pretrained or custom
You decide whether a pretrained model is enough or whether a custom model trained on labelled data is needed.
An accounts team receives thousands of scanned supplier invoices and needs the vendor name, total and due date as structured fields. Vision text detection would return raw text lines but not labelled fields. OCI Language works on text that is already extracted. Document Understanding's key-value extraction for invoices returns the named fields directly, so it is the service the question expects.
Objectives 4
Official objectives from the Oracle outline, in outline order.
7.1 Explore OCI AI services and related APIs: OCI LanguageText analysis, classification, PII detection and translation.
- What it means
- OCI Language applies pretrained natural language processing models to text. Features include language detection, sentiment analysis, named entity recognition, key phrase extraction, text classification, personal identifiable information detection and translation.
- What to understand
- Sentiment can be returned for a whole document, for each sentence, or per aspect, such as positive on price but negative on delivery. Named entity recognition labels people, places, organisations, dates and similar items. PII detection finds and can mask personal data before text is stored or shared. Custom models can be trained for classification and entity extraction. Calls work one record at a time or in batch through jobs on Object Storage.
- How it shows up
- Expect questions that match a text goal, such as masking card numbers or scoring reviews, to a Language feature.
- Common confusion
- Sending audio files to Language; recordings must be transcribed by Speech first.
7.2 Explore OCI AI services and related APIs: OCI VisionImage classification, object detection, text and face detection.
- What it means
- OCI Vision analyses images. Image classification assigns labels to the whole image, object detection finds and locates items with bounding boxes, text detection reads printed or handwritten words in a scene, and face detection locates faces.
- What to understand
- Pretrained models cover common objects and scenes. For domain-specific needs, such as recognising a company's own products or defects, you label images with OCI Data Labeling and train a custom classification or detection model in Vision without writing training code. Classification answers what is in the image; detection also says where. Vision suits photos and video frames, while structured business documents are better served by Document Understanding.
- How it shows up
- A question may ask whether classification or object detection is needed when location matters.
- Common confusion
- Choosing image classification when the task needs to count or locate individual items.
7.3 Explore OCI AI services and related APIs: OCI Document UnderstandingText, table and key-value extraction from documents.
- What it means
- OCI Document Understanding processes documents such as PDFs and scanned images. It extracts text with layout, pulls out tables, returns key-value pairs and classifies the document type.
- What to understand
- Pretrained key-value models exist for common documents including receipts, invoices, passports and driver licences, returning fields such as total or document number with confidence scores. Document classification sorts incoming files, for example separating contracts from invoices. Custom models can be trained for organisation-specific forms. The service suits back-office automation such as accounts payable and onboarding, where output must feed other systems as structured data.
- How it shows up
- Scenarios involving forms, receipts or invoices with named fields usually point to this service.
- Common confusion
- Picking Vision text detection when the requirement is labelled fields rather than raw text.
7.4 Explore OCI AI services and related APIs: OCI SpeechTranscription of audio, plus text-to-speech.
- What it means
- OCI Speech converts spoken audio into text using automatic speech recognition. It returns transcripts with word timestamps and confidence scores, and it can also synthesise speech from text.
- What to understand
- Batch transcription jobs read audio or video files from Object Storage and write results back, while real-time transcription handles live streams. Options include profanity filtering, punctuation, multiple languages and subtitle-friendly output formats. Transcripts often feed OCI Language for sentiment or entity analysis, which is a common two-service pattern in contact centre scenarios.
- How it shows up
- Expect a call recording or meeting audio scenario that requires a transcript before any text analysis.
- Common confusion
- Expecting Speech to return sentiment; that analysis comes from Language after transcription.
High-yield concepts
The terms, rules and formulas this domain turns into questions most often.
-
Aspect-based sentiment
Sentiment scored per topic mentioned in the text.
- Remember
- Different aspects can have different sentiment.
- Tested as
- Review analysis scenarios.
- Trap
- Assuming only one overall score is available.
-
Named entity recognition
Labels people, places, organisations, dates and more.
- Remember
- Feature of OCI Language.
- Tested as
- Feature identification.
- Trap
- Confusing it with key phrase extraction.
-
Key phrase extraction
Pulls out the main phrases in a text.
- Remember
- Summarises topics, not entity types.
- Tested as
- Feature identification.
- Trap
- Treating phrases as typed entities.
-
PII detection
Finds personal data such as names and card numbers in text.
- Remember
- Can mask before storage.
- Tested as
- Privacy scenarios.
- Trap
- Looking for it in Vision.
-
Text classification
Assigns a category to a document.
- Remember
- Pretrained or custom.
- Tested as
- Routing scenarios.
- Trap
- Confusing it with document classification of scanned files.
-
Image classification
Labels the whole image.
- Remember
- Answers what, not where.
- Tested as
- Vision feature choice.
- Trap
- Using it to count objects.
-
Object detection
Finds and locates objects with bounding boxes.
- Remember
- Answers what and where.
- Tested as
- Vision feature choice.
- Trap
- Picking classification when location matters.
-
Custom Vision model
Model trained on your labelled images.
- Remember
- No training code required.
- Tested as
- Domain-specific image tasks.
- Trap
- Assuming pretrained models know your products.
-
Key-value extraction
Returns named fields such as total or date.
- Remember
- Pretrained for receipts, invoices and IDs.
- Tested as
- Document automation scenarios.
- Trap
- Choosing OCR text output instead.
-
Table extraction
Recovers rows and columns from documents.
- Remember
- Preserves structure.
- Tested as
- Financial document scenarios.
- Trap
- Expecting plain OCR to keep table layout.
-
Batch transcription
Speech jobs over files in Object Storage.
- Remember
- Results written back to storage.
- Tested as
- Call recording scenarios.
- Trap
- Expecting instant streaming results.
-
Speech plus Language
Transcribe audio, then analyse text.
- Remember
- Common two-step pattern.
- Tested as
- Contact centre scenarios.
- Trap
- Expecting one service to do both.
Side by side
Concepts candidates mix up, compared on the point that decides the answer.
Which service for which input
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| OCI Language | Analyses text. | Input is already text. | Sending scans or audio directly. |
| OCI Vision | Analyses photos and images. | Scene-level understanding, objects, faces. | Using it for structured form fields. |
| OCI Document Understanding | Analyses business documents. | Fields, tables and document types. | Using it on general photos. |
| OCI Speech | Converts audio to text. | Input is sound. | Expecting sentiment or entities from it. |
Reading text from images
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Vision text detection | Finds words in a photo or scene. | Raw lines and positions. | Expecting labelled fields. |
| Document text extraction | OCR with layout on documents. | Keeps document structure. | Assuming it detects objects. |
| Document key-value extraction | Named fields from known document types. | Structured output for systems. | Using generic OCR then hand parsing. |
Vision feature choice
| Concept | Meaning | Key difference | Common trap |
|---|---|---|---|
| Image classification | Label for the whole image. | No location information. | Using it to count items. |
| Object detection | Labels plus bounding boxes. | Locates each item. | Choosing it when only a yes or no label is needed. |
Exam traps
Each trap is an answer that looks right. Learn the clue that should change your mind.
-
Sentiment from audio
- What candidates pick
- OCI Speech for call sentiment
- Why it looks right
- The input is audio, so Speech seems complete.
- Why it's wrong
- Speech produces the transcript; sentiment analysis is an OCI Language feature applied to that text afterwards.
Clue to watch for Both audio input and an opinion-based output.
-
OCR versus fields
- What candidates pick
- Vision text detection for receipts
- Why it looks right
- Vision reads text in images.
- Why it's wrong
- Text detection returns words and positions, while Document Understanding key-value extraction returns named fields like totals.
Clue to watch for Named fields or structured output.
-
Classification versus detection
- What candidates pick
- Image classification for counting cars
- Why it looks right
- It is the simplest Vision feature.
- Why it's wrong
- Classification gives a label for the whole image and cannot locate or count separate objects.
Clue to watch for Words like locate, count or where.
-
Entities versus phrases
- What candidates pick
- Key phrase extraction to find company names
- Why it looks right
- Both pull terms from text.
- Why it's wrong
- Named entity recognition labels organisations, people and places by type, while key phrases are untyped topics.
Clue to watch for Requirement to know what type each term is.
-
Pretrained knows everything
- What candidates pick
- Pretrained Vision for proprietary parts
- Why it looks right
- Pretrained models seem broad.
- Why it's wrong
- Pretrained models cover common objects; recognising specific in-house items needs a custom model trained on labelled images.
Clue to watch for Company-specific products or defects.
-
Document type routing
- What candidates pick
- OCI Language text classification for scanned files
- Why it looks right
- Classification sounds like the same feature.
- Why it's wrong
- Scanned files need Document Understanding, whose document classification works on the document image directly.
Clue to watch for Incoming PDFs or scans to sort.
Answer framework
Use these steps to pick the right pretrained OCI AI service for a Domain 7 scenario.
- Identify the input format: plain text, a photo or video frame, a scanned business document or PDF, or recorded or live audio.
- Map the input to a starting service: Language for text, Vision for images, Document Understanding for documents and Speech for audio.
- Pin down the exact output: sentiment, entities, PII, labels, bounding boxes, key-value fields, tables or a transcript.
- If the task chains steps, such as audio to transcript to sentiment, select both services in order rather than one.
- Decide whether pretrained models cover the content or whether a custom model trained on labelled data is needed, such as custom Vision object detection for a company's own product images.
Practice question
Want more like this? The 1Z0-1122-26 practice test has questions for every domain.
Checklist
0 of 11
Tick each one you can do without notes. Your ticks stay in this browser.
Study strategy
Give this section six to eight hours, since it likely carries the most questions. Watch the AI services lessons in Oracle's course, then read the overview page for each of the four services in the OCI documentation and list every feature on a card per service.
If you have console access, try each service once. Paste a review into Language and look at aspect sentiment and entities. Upload a photo to Vision and compare classification with object detection. Upload a receipt to Document Understanding and see the key-value output. Even ten minutes per service makes features far easier to recall.
Then build an input-to-service matrix: rows for text, image, document and audio, columns for typical outputs. Practise with scenario questions, naming the input type before reading the options.
Pay special attention to overlap pairs: Vision text detection versus Document Understanding, and Speech versus Language. Most errors in this section come from those two boundaries, so review them again the day before the exam.
FAQ
How many questions come from Intro to OCI AI Services?
Oracle does not publish weights for 1Z0-1122-26. Secondary estimates make this OCI AI services section the largest, around eight of the 40 questions.
Which services are in scope?
Language, Vision, Document Understanding and Speech are named in the objective. Other services such as Digital Assistant appear mainly in the portfolio section.
Do I need to know API syntax?
No. Know what each service accepts and returns and how it is accessed through console, REST APIs and SDKs.
Can these services be customised?
Yes. Language, Vision and Document Understanding support custom models trained on your labelled data, while pretrained models handle common cases.
What is the most common error in this section?
Mixing up Vision text detection with Document Understanding, or expecting Speech to provide sentiment.
How do Speech and Language work together?
Speech transcribes audio into text, and Language then analyses that text for sentiment, entities or PII.
Why does Oracle list only one objective here?
Secondary sources show a single objective covering the services and their APIs. This guide splits it into four topics, one per service, for easier study.
Other Artificial Intelligence & Machine Learning certifications
Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.
Test your Domain 7 knowledge
You have reviewed OCI Language, Vision, Document Understanding and Speech. Test how quickly you can match an input and goal to the right service.
Keep going
- Training Training Built on your Weak areas
- Practice test Questions by domain, with explanations
- Mock exam Full length, timed like the real exam
- Readiness test Find your weakest domain in minutes
- Cheat sheet Key terms and formulas on one page
- Ask the 1Z0-1122-26 AI tutor Get any concept explained your way