Using Corpus Evidence to Choose a Verb Form | Filipino Grammar

A corpus—a searchable collection of real spoken or written language—can answer questions that an affix chart cannot. It can show whether speakers actually use a predicted form, which markers surround it, what meaning recurs, and whether it belongs mainly to conversation, news, academic prose, a region, or an earlier period. It cannot, by raw frequency alone, declare a form grammatical or choose the best sentence for your context.

Filipino makes corpus work especially valuable because one root supports many voice and aspect families, surface forms can be ambiguous, and productive-looking affixes have lexical restrictions. The method on this page treats corpus evidence as usage evidence to be interpreted alongside a reliable dictionary and grammatical analysis. Begin with the principles on choosing a dictionary citation form, then use corpora to investigate the larger paradigm.

What to Search—and What It Can Show

Research question	Search strategy	Evidence to record	What not to conclude
Is a form established?	search the exact surface form and nearby paradigm members	independent tokens across speakers and texts	one web hit proves productivity
Which voice frame occurs?	inspect nearby ang, ng, sa, si, ni, kay	roles and markers in complete clauses	English word order identifies the pivot
Which aspect is intended?	search completed, incompleted, and contemplated forms	event viewpoint plus temporal context	one form equals one English tense
Is a variant standard?	compare edited, conversational, regional, and code-switched sources	genre, date, region, and editorial status	the most numerous spelling is always preferred
What does an ambiguous form mean?	read a full concordance window	arguments, collocates, and discourse topic	the isolated word has one root

The University of the Philippines' Filipino Language Corpus project is useful as a model because its design recognizes both spoken and written data and distinguishes genres. Any corpus you use should be evaluated similarly: who produced the texts, when, in what setting, and with how much editing?

Start with a Root Family, Not an English Word

Suppose you want to express English cook. Searching only cook = luto hides the choices among nagluto, niluto, ipinagluto, pinaglutuan, and other lexical families. Begin with a dictionary-supported root, list plausible surface forms, and search complete contexts.

Nagluto si Mara ng ginataang gulay para sa mga bisita.

Mara cooked vegetables in coconut milk for the guests. (actor voice; neutral)

Niluto ni Mara ang ginataang gulay bago dumating ang mga bisita.

Mara cooked the vegetables in coconut milk before the guests arrived. (patient voice; neutral)

These controlled clauses describe similar events, but the corpus frames differ. Nagluto frequently supports an ang-marked actor and ng-marked food; niluto supports an ng-marked actor and ang-marked food. A search for the verb alone is therefore incomplete—record the marker frame.

💡

Build a small paradigm before searching: root, candidate derivation, aspect forms, and likely markers. A corpus search should test an analysis, not replace one.

Search More Than the Citation Form

Dictionaries often cite a basic or contemplated-looking form, but running text may favor its completed or incompleted members. For makipag-usap, search at least makipag-usap, nakipag-usap, nakikipag-usap, and makikipag-usap. Include standard hyphenation, and examine snippets only after opening the full context.

Nakipag-usap si Tomas kay Aling Fe tungkol sa butas na bubong.

Tomas spoke with Aling Fe about the hole in the roof. (completed; neutral)

Nakikipag-usap si Tomas sa kontratista tuwing may bagong sira.

Tomas talks with the contractor whenever new damage appears. (incompleted habitual; neutral)

If the completed form dominates a small sample, that does not mean the incompleted form is defective. Your corpus may contain more news reports than live conversation, or your topic may favor bounded events.

Search spelling variants cautiously. Normalize only for counting; preserve the attested spelling in your notes, then check it against standard orthography. Search-engine result totals are unstable and opaque, so treat ordinary web search as a discovery tool, not a balanced corpus.

Read Concordance Windows for Meaning and Valency

A concordance displays repeated occurrences with surrounding words. Read enough context to identify participants and meaning. Surface forms can hide different roots or categories.

Namimili si Lorna ng sariwang isda tuwing madaling-araw.

Lorna shops for fresh fish every dawn. (bili family; neutral)

Namimili si Lorna ng pinakamurang pamasahe bago bumiyahe.

Lorna chooses the cheapest fare before traveling. (pili family; neutral)

Both clauses contain namimili, but the first belongs to bili “buy/shop” and the second to pili “choose.” Collocates such as the selected alternatives and the wider situation disambiguate them. Counting the word as one lemma would corrupt the result.

The same problem crosses word classes:

Mananahi si Elsa ng mga kurtina kapag dumating ang tela.

Elsa will sew the curtains when the fabric arrives. (contemplated verb; neutral)

Isang mananahing kilala sa buong bayan si Elsa.

Elsa is a dressmaker known throughout the town. (noun predicate; neutral)

The same ordinary spelling can represent a verb or a profession noun. Marker patterns and following complements—not the string alone—identify the analysis.

Compare Genres Before Choosing a Form

Frequency is always frequency in a dataset. A news-heavy corpus favors formal institutional verbs; conversational transcripts reveal particles, omissions, and code-switching; literary texts preserve stylized or older forms. Divide evidence by genre, date, region, and mode whenever metadata permits.

For example, magsagawa ng pagsusuri is productive in (formal/academic/administrative) prose. In conversation, speakers often prefer suriin or another simpler verb. Nagsiuwian remains familiar, but other magsi- formations may sound (formal), (literary), or deliberately playful. A corpus that mixes genres without labels conceals those facts.

Magsasagawa ng pagsusuri ang pangkat sa kalidad ng tubig.

The team will conduct an analysis of the water quality. (formal/technical)

Susuriin ng pangkat ang tubig mula sa tatlong balon.

The team will examine the water from three wells. (plain formal-neutral)

Neither sentence is universally “more Filipino.” The first packages an institutional activity as magsagawa + event noun; the second uses a direct patient-voice verb. Genre and desired information structure decide between them.

💡

A high count may reflect one publisher, one copied press release, or one prolific author. Count independent sources and inspect dispersion, not just tokens.

Test Productivity with Restraint

An affix is productive when speakers can extend it to appropriate new bases, but corpus absence does not automatically prove impossibility. Rare events, new terms, and oral forms may be missing. Conversely, one creative token does not prove unrestricted productivity.

Use a three-way judgment:

Established: supported by dictionaries and dispersed contemporary corpus examples.
Possible but marked: interpretable and attested, but concentrated in a register, region, genre, or creative context.
Unsupported for active use: predicted only by an affix formula, with no reliable contemporary evidence.

This method is especially important for nasal substitution, reciprocal derivation, and plural-action forms. It prevents both prescriptive erasure of real variation and mechanical invention.

Hindi namin ginamit ang bagong anyo dahil iisa lamang ang nakita naming halimbawa nito, at iyon ay isang komentong hindi sinuri ng patnugot.

We did not use the new form because we found only one example of it, and it was an unedited comment. (neutral research context)

The sentence models an evidential decision, not a grammatical ban. A larger spoken corpus could change the classification.

A Reproducible Workflow

For each candidate form, keep a small record:

Identify the root and proposed affix family in a reliable dictionary or grammar.
Generate only morphologically justified aspect forms.
Search exact forms, then search likely marker frames and collocates.
Remove duplicates, quotations of the same source, scanning errors, and irrelevant homographs.
Annotate voice, aspect, participant markers, meaning, genre, date, and region.
Compare a competing construction built from the same root or meaning.
Write a conclusion that states both the evidence and its limits.

—Alin ang mas karaniwan sa usapan, magsagawa ng pagsusuri o suriin? —Ihiwalay muna natin ang pasalita at nakasulat na datos bago magbilang.

—Which is more common in conversation, magsagawa ng pagsusuri or suriin? —Let's separate the spoken and written data before counting. (academic discussion)

This exchange captures the core discipline: define the relevant population before comparing frequency.

English Interference in Corpus Interpretation

English speakers often search only the translation they want, then copy the first Filipino verb beside it. That reverses the evidence. Start from Filipino forms and their marker frames. Never add -ed, -ing, or -s to manufacture a missing cell, and never assume one affix has one predictable English gloss.

A root is a construction family. Corpus work helps reveal that family, but the analyst must still separate aspect, voice, derivation, valency, register, and lexical meaning.

Common Mistakes

Treating one hit as proof of a productive form

❌ Gamitin natin agad ang anyo dahil nakita ito sa isang komento.

Poor method—the single unedited token may be a typo, innovation, or idiosyncratic use.

✅ Maghanap muna tayo ng magkakahiwalay na halimbawa at suriin ang buong pangungusap.

Let's first find independent examples and examine the complete sentence. (neutral research advice)

Counting homographs as one verb

❌ Iisa ang kahulugan ng lahat ng halimbawa ng mananahi.

Incorrect—the spelling can be a contemplated verb or a profession noun.

✅ Paghiwalayin natin ang mananahing pandiwa at ang mananahing pangngalan.

Let's separate verbal mananahi from nominal mananahi. (academic/technical instruction)

Ignoring marker frames

❌ Pareho lang ang nagluto at niluto dahil iisa ang salin nila sa Ingles.

Incorrect analysis—the forms may share an English translation but select different pivots and marker frames.

✅ Ihambing ang nagluto si Mara ng gulay at niluto ni Mara ang gulay.

Compare nagluto si Mara ng gulay with niluto ni Mara ang gulay. (metalinguistic instruction)

Letting raw counts erase register

❌ Pinakamainam sa lahat ng sitwasyon ang anyong pinakamarami sa mga ulat ng pamahalaan.

Incorrect inference—a government-heavy sample cannot represent every conversational setting.

✅ Tukuying pormal ang anyo kung karamihan sa mga halimbawa nito ay nasa mga dokumentong opisyal.

Identify a form as formal when most of its examples occur in official documents. (academic advice)

Inventing an English-style paradigm

❌ Nagluto-ed ang anyong ginamit para sa natapos na pangyayari.

Incorrect—English -ed is not evidence and never attaches to Filipino aspect forms.

✅ Hanapin ang nagluto bilang anyong may pokus sa tagaganap at nagsasaad ng natapos na pangyayari.

Search for nagluto as the completed actor-voice form. (metalinguistic instruction)

Key Takeaways

Search paradigms and marker frames, not isolated English translations.
Read full concordance contexts to separate roots, meanings, word classes, voice, and valency.
Compare genre, date, mode, region, and source dispersion before making a usage claim.
Frequency supports a register-sensitive conclusion; it does not by itself define grammaticality.
Triangulate corpus evidence with dictionaries, grammars, orthographic standards, and careful speaker judgment.

Now practice Filipino

Reading grammar gets you part of the way. The exercises are where it sticks — free, no signup needed.

Start learning Filipino →